|
Title: |
ANN-ABC META-HEURISTIC HYPER PARAMETER TUNING FOR MAMMOGRAM CLASSIFICATION |
|
Author: |
AJAY KUMAR MAMINDLA, DR. Y. RAMADEVI |
| Abstract: | In recent past, artificial neural networks (ANN) have reaped improvements in the domain of medical image processing by addressing many unmanageable problems. The initialized hyperparameters control ANN performance and selecting sensible hyperparameters by hand is time-consuming and tiresome. This study suggests a metaheuristic optimization of the fine-tuning hyperparameters approach to remedy this flaw. The method is then evaluated on mammography images to assess whether the mammogram contains cancer. In the proposed ANN model, a modified Artificial bee colony (ABC) optimization method is used to fine tune the hyperparameters, and it categorizes the tumors in the breast as benign or malignant in two-class case and normal, benign, and malignant in three-class case with an accuracy of 97.52% and 96.58% respectively. Hyperparameters to the neural network framework were assigned instantly with the help of ABC method with wrapped ANN as objective function. Manual search, Grid Search, Random Grid search, Bayes search are all cutting edge ANN hyperparameters methods. In addition to the mentioned, nature-inspired optimization methods such as PSO and GA have adopted for fine tuning parameters. Additionally, the suggested model's performance in classifying breast pictures was compared to that of the published hyperparameter technique using sizable datasets on breast cancer that were made accessible to the public. |
|
Keywords: |
Artificial Neural Networks, Hyperparameters, Artificial bee colony, Mammogram images, Grid Search. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
DEVELOPMENT OF MATHEMATICAL MODELS TO DETERMINE THE BALANCE OF THE SYSTEM OF PLATFORM INTERACTIONS WHEN SCALING THE END-TO-END MONITORING PROCESS FOR PRIORITY SECTORS OF THE ECONOMY |
|
Author: |
OLGA VALERYEVNA SEDOVA , ALEXEY GRIGORIEVICH ALEKSEEV |
| Abstract: | Prerequisites for the present study are the emergence and development of new forms of business cooperation in the process of digitalization of society in the form of digital platforms. In scaling digital solutions when entering new markets of relevance is the issue of preserving the balance of the platform interactions system. In view of the complexity of such systems due to the variety of objects and forms of platform interaction, the goal of the study is to develop tools for determining the degree of balance of the platform interactions system in scaling the end-to-end monitoring process for the priority sectors of the economy in the form of mathematical models. The employed research methods include analysis, the parametric method, modeling, and methods of value engineering. The study proposes to consider the presence of a synergistic effect in the assessment of the balance of the system of platform interactions in scaling the end-to-end monitoring process for the priority sectors of the economy. The effect of types of end-to-end monitoring processes on the value of the synergistic effect is considered and a mathematical model for its calculation is proposed. To make informed decisions when scaling digital platforms to other priority sectors of the economy, a mathematical model for determining the imbalance zones of platform interaction is proposed. Considering the performance of the end-to-end monitoring process when scaling digital platforms to determine the flexibility factor, a mathematical model for calculating the design efficiency indicator is offered. |
|
Keywords: |
Platform Interaction, End-to-End Process, Monitoring, Mathematical Model, Synergy. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
A SYSTEMATIC LITERATURE REVIEW OF DEEP AND MACHINE LEARNING ALGORITHMS IN BRAIN TUMOR AND META-ANALYSIS |
|
Author: |
ASHRAF M. H. TAHA, DR. SYAIBA BALQISH BINTI ARIFFIN, SAMY S. ABU-NASER |
| Abstract: | Brain tumor (BT) is considered one of the dangerous conditions that could strike both adults and children. 85 to 90% of all primary malignancies of the “Central Nervous System (CNS)” are Brain Tumor. Each year, brain tumors are discovered in approximately 11,700 persons. The 5-year survival rate for patients with malignant brain or CNS tumors is around 36% for women and 34% for men. The current systematic review depends on “the Preferred Reporting Items for Systematic reviews and Meta-Analysis statement” and 40 appropriate studies. The search of the literature employed search engines similar to: IEEE Xplore, Google Scholar, Hindawi, PubMed, SCOPUS, Wiley Online, Web of Science, Taylor and Francis, Science Direct, and Ebscohost. This study concentrated on four characteristics: Algorithms of Machine and Deep Learning, best- algorithm performance, datasets, and application used in Brain Tumor predictions. The experimental articles did not use Reinforcement Learning, Semi-supervised learning, and promising aspects of Deep and Machine Learning. Algorithms based on ensemble technique exhibited sensible rates of accuracy nonetheless were not frequent, whereas Convolutional Neural Network (CNN) were well epitomized. A few studies smeared main datasets (13 of 40). Logistic Regression (LR), Deep Neural Network (DNN), boosting algorithms, Support Vector Machine (SVM) and K-Nearest Neighbors (KNN), were the best performing algorithms. This review will be beneficial for investigators predicting Brain Tumor using machine and deep learning methods. |
|
Keywords: |
Brain Tumor, Datasets, Deep Learning, Machine Learning |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
A NEW CONTRIBUTION OF IMAGE ENCRYPTION BASED ON CHAOTIC MAPS AND THE Z/nZ GROUP |
|
Author: |
ELAZZABY FOUZIA, ELAKKAD NABIL, SABOUR KHALID, KABBAJ SAMIR |
| Abstract: | Focusing on a significant scientific advancement in image encryption, this paper is an excellent example of its kind. It makes use of a 2D sinusoidal logistic modulation map and the advantageous transformation features of the Z/nZ group . in witch, we have developed a more suitable key flow by which to create our Z/nZ groups as a result of the exceptional hyper-chaotic and ergodic features of our 2D-SLMM maps. By randomly rearranging the orientation of its pixels, it is possible to generate a blurring pattern that has no foreseeable relationship to the original image. Because of this, the transmitted image is no longer recognizable as being based on the original, which is now blurred and unreadable, and its transmission is private and safe from prying eyes. In fact, we compared our algorithm to five different methods from the literature and employed metrics including the histogram, entropy, correlation analysis, and differential assaults. Our simulation findings show that it performs well and achieves a high level of security with the optimum algorithmic complexity and adequate protection against unlawful manipulation. In other words, when compared to alternative methods, our method performs well when encrypting images. |
|
Keywords: |
Image Encryption, Group Z/nZ, Chaotic Maps, 2D-SLMM And Security |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
MULTI-CLASS WEATHER CLASSIFICATION USING MACHINE LEARNING TECHNIQUES |
|
Author: |
ADOLF FENYI, MICHAEL ASANTE |
| Abstract: |
In the field of computer vision, multiclass weather classification from outdoor
images is a difficult task to perform due to the diversity and lack of unique
weather features. In this paper, a novel algorithm is formulated using machine
learning techniques to classify several weather conditions such as sunny,
cloudy, rainy, snowy and hazy. The researchers implemented a novel edge
detection algorithm for the segmentation purpose while the Support Vector
Machine was used for the classification task. However, before any classification
is done, several weather features such as sky, cloud, rain streaks, snowflakes
and dark channels are extracted from segmented images to increase the efficiency
of the classifier. The extracted features are later concatenated and read into
SVM for training and classification purposes. The experiment revealed that multi-feature concatenation is essential since it yielded an average performance of 80.4% as compared to single feature selection of 72.8%. From the evaluation of the proposed algorithm with other recent weather classification algorithms, the proposed algorithm exhibited an accuracy and a time complexity of (80.4%, O(n2)) against the XGBoost (Extreme Gradient Boosting algorithm - 70.5%, O(n2)), MLRA (Multilevel Recognition Algorithm - 75.1%, O(n3)) and CNN (Convolution Neural Network - 83.9%, O(n4)). |
|
Keywords: |
Edge detection; Weather classification; Support Vector Machine; Convolution Network Architecture; Object segmentation |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
ENERGY EFFICIENT BASED OPTIMIZED K-MEANS AND MODIFIED WHALE OPTIMIZATION ALGORITHM FOR CLUSTER HEAD SELECTION IN WSN |
|
Author: |
S.KAVIARASAN, R. SRINIVASAN |
| Abstract: |
Background: More low-cost and power sensor nodes make up a wireless sensor
network (WSN). Using self-organization, the sensor nodes build a wireless
network in a specific area. In spite of the fact that people can't close the
doors, they can still function normally in any special or wicked environment.
Due to a variety of complex factors, effective data transmission between nodes
is nearly impossible. When it comes to improving the efficiency of data
transmission, clustering is a well-known strategy. The clustering model splits
sensor nodes into different groupings. All sensor nodes in a cluster get
information from a single cluster head node. Objective: To improve the network lifetime, throughput and reduce energy consumption by using K-means and modified whale optimization algorithm (KM-MWOA). Methods: Clustering algorithms play an important role in selecting the most energy-efficient and least-delayed cluster head under such conditions. K-means algorithm is used to select cluster head selection (CHS) and then present the modified whale optimization algorithm (MWOA) to convey packets in multi-hop transmission among CHS and the BS and choose an optimal path. During the global search phase, random population seeding is used to increase the standard WOA. Algorithms are able to discover in the early stages of the search, while also utilizing the search space lengthily in the later stages, by changing the parameters A and b. Reduced intra-cluster communication and improved energy efficiency for sensor nodes are both possible with this clustering scheme. Result: The KM-MWOA strategy has been realised performance metrics such as Network lifetime, Energy Consumption, Network throughput in MATLAB R2018a with NVIDIA Tesla K80 GPU. The results of the KM-MWOA is compared with Radio Access Technologies, Jellyfish Algorithm. Conclusion: The KM-MWOA is proposed for extending the lifetime of sensor nodes, reducing the energy consumption and improving throughput. Energy Efficient is diagnosed by using the KM-MWOA models. It is concluded that extending the power of the KM-MWOA supersedes all previous methods (Radio Access Technologies, Jellyfish Algorithm). |
|
Keywords: |
K-Means Algorithm, Modified Whale Optimization Algorithm, Wireless Sensor Network, Energy Consumption; Optimal Cluster Head |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
LINK STABILITY AND OBSTACLE AVOIDANCE BASED TRAFFIC-AWARE REACTIVE ROUTING PROTOCOL FOR MOBILE AD-HOC NETWORK |
|
Author: |
VEERAMANI. R, DR. R. MADHANMOHAN, DR.C.MAHESH |
| Abstract: | A distinctive type of wireless network called an ad hoc network has several hops, no central hub, and dynamic architecture. The forwarding of intermediate nodes is required to achieve data communication. A mobile ad hoc network is made up of several wireless communication nodes, each of which serves as a router. The restrictive characteristics of this type of communication networks, such as its frequent topological changes and low battery power, have a significant impact on routing. Numerous studies have been conducted to enhance MANET's routing performance. The network's performance typically suffers as a result of the unstable path caused by the high traffic load on links and the low energy of the nodes. Reactive MANET routing techniques handle link breaks caused by node mobility and energy consumption. For broad area networks, multi-hop communication may be more effective in terms of routing methods. To build a better quality route between source and destination, the research suggested the Traffic-Aware and Stable Ad hoc On-Demand Multipath Distance Vector protocol (TAS-AOMDV), which is an upgrade over AODV. By choosing the most stable neighbor concerning the transmitter of the route request message and the nodes approaching throughout the route discovery process, this method aims to forecast a stable path. To reduce lost data packets and route error messages, this minimizes the contention phase, predicts the route lifespan, and speeds up data packet transmission. The effectiveness of routine procedures in MANETs is substantially hampered by the existence of obstructing obstacles. This study instead looks at the dynamic and autonomous detection of potential network barriers. A combined Lion Optimization algorithm (LOA) and Dynamic Window Approach are proposed in the article to suit the needs of global optimal and dynamic obstacle avoidance in path planning (DWA). To find the path, this optimization technique has been used. The algorithm proficiently plots a path among multiple nodes over the plot area, which results in perceiving the shortest path that is optimal without obstacles. The performance analysis was carried out in Network Simulator 3.36 software. It took into account various aspects like packet delivery ratio end-to-end delay for differing traffic loads, throughput and in the proposed scheme via existing schemes. |
|
Keywords: |
MANET, Routing Protocol, Traffic Loads, AOMDV, Obstacle Avoidance, Lion Optimization Algorithm, Dynamic-Window Approach, Network Simulator. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
ANALYSIS OF FACTORS AFFECTING ADOPTION OF CLOUD ACCOUNTING IN INDONESIA |
|
Author: |
SULINA ZEBUA, RINDANG WIDURI |
| Abstract: | Cloud accounting is a novelty in accounting information systems. Cloud accounting is the result of the digitization process in accounting, based initially on complex traditional application systems to cloud-based applications to handle accounting tasks more flexibly and efficiently. Research on cloud accounting has not become a common topic in developing countries, especially in Indonesia. This study analyzes the factors of cloud accounting in Indonesia. A self-administered questionnaire was conducted with 175 accounting staff in Indonesia. The results indicate that top management support, organizational competency, service quality, and system quality positively affect the perceived usefulness and ease of use of cloud accounting. Perceived usefulness positively affects the intended use, while perceived ease of use positively impacts the perceived usefulness and intention to use cloud accounting. Intention to use cloud accounting has a positive effect on adoption. Therefore, accounting staff can adapt to the dynamic technology innovation by investing in cloud accounting, which has the potential for high industry values. The use of cloud accounting can also facilitate accounting staff to manage their work better. The ability of the small firm to use cloud accounting is a crucial factor in creating new development for industry continuity, accompanied by establishing relationships based on data. |
|
Keywords: |
Cloud Accounting, TOE (Technology, Organization, Environment), TAM (Technology Acceptance Model), D&M (DeLone-McLean) |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
CUCKOO SEARCH SUPPORT VECTOR MACHINE FOR SUPPLY CHAIN RISK MANAGEMENT |
|
Author: |
P.RAMAN, R.SEETHA, S.SANKAR, K.SURESH, R.ARUNKUMAR, T.A.MOHANAPRAKASH |
| Abstract: | Supply chain interruptions have been identified as a key risk factor. Supply chain risk management has been driven by technological advancements, an increase in information overload, and a greater exposure to risk. Data mining uses a variety of analytical methods to make intelligent and fast decisions; yet, its utility in supply chain risk management has yet to be fully realized. The risk in the supply chain is prioritized using machine learning techniques. The majority of supply chain studies, on the other hand, focus on prediction efficiency and ignore the significance of interpretability, which helps experts to mitigate or avoid risks. The goal of this study is to develop a data mining-based cuckoo search support vector machine supply chain risk management (DM-CSVM SCRM) for predicting hazards in supply chains, as well as identifying, assessing, and mitigating them. A supply chain risk prediction is done by using a machine learning algorithm in this project. A holistic approach to risk management combines risk management and DM operations into a single structure for efficient management of risk. Based on discussions, focus group interviews, and semi-structured interviews the framework is tested in this study. The research shows how DM can help you find unseen and relevant data in unstructured risk data so you can make better risk decisions. |
|
Keywords: |
Supply Chain Risk Management, Decision Support System, Data Mining, Machine Learning, Data Analytics. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
A GRID CONNECTED HYBRID RENEWABLE ENERGY SYSTEM FOR OPTIMAL ENERGY MANAGMENT BASED ON ANT-LION OPTIMIZATION ALGORITHM |
|
Author: |
IBRAHIM A. ALTAWIL , KHALED A. MAHAFZAH , AMMAR ALMOMANI |
| Abstract: | Over the past several years, the need of hybrid renewable energy systems (HRES) integrated with the electrical grid system has significantly increased. This integration provides a better reliability, continuous supply, and improved system performance. The optimal size (rating) of a grid connected HRES is proposed in this paper. The Ant-Lion Optimization (ALO) algorithm is used for optimization. For the ALO algorithm, the decision variables are the number of PV panels (NPV) and wind turbines (NWT). The optimization considers a particular priority based on limitations to supply the load demand connected with the hybrid system from PV arrays, wind turbines, and ultimately the grid. Additionally, the impact of independently combining solar PV and wind with a single objective function is examined. The Total Net Present Cost (TNPC) and the Index of Reliability (IR), which serve as the objective function, are minimized using the ALO. Two different scenarios are proposed. After running the optimization algorithm, the two proposed scenarios are compared with each other in term of TNPC and IR. Thus, this determines the number of PVs and wind turbines to build HRES. |
|
Keywords: |
Hybrid Renewable Energy Sources (HRES), Load Demand, Electrical Grid, Ant-Lion Optimizer, Energy Cost |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
THE IMPACT OF THE NATIONAL ECONOMY DIGITALIZATION ON THE EFFICIENCY OF THE LOGISTICS ACTIVITIES MANAGEMENT OF THE ENTERPRISE IN THE CONDITIONS OF INTENSIFYING INTERNATIONAL COMPETITION |
|
Author: |
OLHA POPELO, SVITLANA TULCHYNSKA, GALYNA KRASOVSKA, OLENA KOSTIUNIK, LARISA RAICHEVA, ОLEKSII MYKHALCHENKO |
| Abstract: |
In the article, the main directions of the logistics activities digitalization
in Ukraine are examined. The purpose of the study is to substantiate the impact
of digitalization on the efficiency of the logistics activities management of
enterprises in the conditions of intensifying international competition. To
achieve the goal, the systematic approach was used, which allows to study the
components of the logistics system as separate elements and their synergistic
effect as a whole. It was determined that for effective operation of transport and logistics companies, the presence of an appropriate level of digital support and the use of software products for the implementation of production operations by enterprises is necessary. It has been proven that the use of modern digital products allows enterprises to take advantage of a single digital global space in the logistics transportation sector. The state of the implementation of digital technologies at logistics enterprises, which are leaders of the industry in Ukraine, is analyzed. The methodical support proposed in the work allows to determine the effectiveness of measures related to the digitalization of the enterprise's logistics activities. Based on the methodological support, the approval of which was carried out on the example of five enterprises of the logistics industry in Ukraine, the development level of the logistics activity of enterprises in the conditions of the production digitalization was established. The conducted calculations showed that for successful activities in the logistics field and activities in the international market, enterprises should implement digital technologies to meet European standards for the relevant services provision. It has been proven that the strengthening of integration processes in the digital technologies sphere in logistics activities contributes to the development of 4PL and 5PL outsourcing, which require a quick exchange of information during the implementation of logistics operations. |
|
Keywords: |
Digitalization, Logistics Activity, National Economy, Management, Enterprise, International Competition, Logistics, Supply Chains |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
A DEEP LEARNING BASED TECHNIQUE FOR THE CLASSIFICATION OF MALWARE IMAGES |
|
Author: |
MD. HARIS UDDIN SHARIF, NASMIN JIWANI, KETAN GUPTA, MEHMOOD ALI MOHAMMED, DR.MERAJ FARHEEN ANSARI |
| Abstract: | Because of the fast expansion of the internet and technology, a slew of developing malware and attack techniques has evolved. As a result, researchers concentrated their efforts on machine learning and deep learning techniques to detect malware. Many organizations have been developing new algorithms and products to secure people from these scams. On the other hand, Malware kinds have been expanding substantially in recent years. The anti-virus companies have been discovering millions of new malware variants every year. Therefore, new intelligent malware detection methods must be solved as soon as possible to halt this rise. Malware is becoming more prevalent, more diverse, and more sophisticated. Deep learning in malware detection through images has recently been demonstrated to be highly effective. We also employed an Image-based Malware dataset [Malimg] and used the different deep learning algorithms, CNN, Caps-Net, VGG16, ResNet, and InceptionV3, for malware detection. The dataset images were transported through the pre-processing pipeline and into the deep learning pipeline, where they were used to train deep learning models in the right way. As part of the model training process, all images were resized to be the same size and proportions. A factor of 1/255 was then applied to the images, resulting in a conversion from RGB value to grayscale, which restored the original RGB values to their correct positions. Later, the dataset was segmented into two groups, train, and test. The VGG16, ResNet50, and InceptionV3 models detected the malware images. A combination of the Adam optimizer and the cross-entropy loss function was used to train all of the models. The models were trained for 50 epochs using early stopping criteria. Finally, the model composition method was used to classify malware images where the previously trained models were combined. The custom CNN model, the VGG16, ResNet50, and InceptionV3 models were combined to predict a single outcome for the experimental condition. The proposed technique provided very promising results. |
|
Keywords: |
Malware Prediction, VGG16, ResNet50, Caps-Net, Image-Based Malware Prediction, Cyber Analysis, Deep Learning, Cyber Security |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
THE EFFECT OF INFORMATION TECHNOLOGY ADOPTION, ENTREPRENEURIAL ORIENTATION ON DYNAMIC CAPABILITIES AND COMPANY PERFORMANCE |
|
Author: |
I NYOMAN SARYA, MTS. ARIEF, HARDIJANTO SAROSO, AGUSTINUS BANDUR |
| Abstract: | The purpose of this study was to determine the effect of information technology adoption, entrepreneurial orientation on dynamic capabilities and the performance of 3, 4, & 5 -star hotel companies in Indonesia. Researchers take a quantitative approach by measuring the sample variables built from the construct and representing the research population. The analysis unit in this study was a company engaged in 3, 4 and 5 -five -starred accommodation services that were spread throughout Indonesia, namely in Bali, West Java, DKI Jakarta, Central Java, East Java, DI Yogyakarta, Banten, Lombok, Sulawesi, Kalimantan, Sumatra, while the Observation Unit is the General Manager of the Hotel. The Partial Least Square (PLS) method is used in this study to analyze responses. The results of this study indicate that Information technology adoption has a positive effect on Dynamic Capabilities, Entrepreneurial orientation has a positive effect on Dynamic Capabilities, Information technology adoption has no effect on Company performance, Entrepreneurial orientation has no effect on Company performance, Dynamic Capabilities have a positive effect on Company performance, Dynamic Capabilities Proven to be able to mediate the relationship of information technology adoption to the Company Performance and Dynamic Capabilities proven to be able to mediate the relationship of information technology adoption to the Company Performance. |
|
Keywords: |
Company Performance, Information Technology Adoption, Dynamic Capabilities, Entrepreneurial Orientation |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
EFFECT OF TECHNOLOGY ORGANIZATION ENVIRONMENT AND INDIVIDUAL FACTORS TOWARDS ADOPTION INTENTION OF CLOUD-BASED ACCOUNTING SOFTWARE IN MSMES |
|
Author: |
ANG SWAT LIN LINDAWATI, BAMBANG LEO HANDOKO, ISABEL JOYCELINE |
| Abstract: | The purpose of this research is to examine the adoption of technology, organization, environment, and individual factors towards the intention of cloud-based accounting software in MSMEs. The research method used is descriptive statistics with survey as data collection method which was distributed to owners/co-owners/manager of micro enterprises in Tangerang Regency. This research uses multiple regression analysis with SPSS version 25 as the data analysis tool. The result of the research shows that technology, organization, environment, and individual factors simultaneously affect the adoption intention of cloud-based accounting software in MSMEs, however technology and environment factors are the only factors that partially affect the adoption intention of cloud-based accounting software in MSMEs. The result contrasted results of the previous research but it is also in line with most research which stated that technology, organization, environment, and individual factors simultaneously affects the adoption intention of technology advancement. |
|
Keywords: |
TOE Framework, Technology, Adoption, Intention, MSME, Micro, Enterprise. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
ANALYZING FACTORS BETWEEN YOUTUBE USAGE AND ADDICTION IN INDONESIA |
|
Author: |
NJOO, DANIEL HARIJANTO , RIYANTO JAYADI |
| Abstract: | YouTube is a social media used often by many people worldwide, including in Indonesia. There are many reasons why many people use YouTube, ranging from personal enjoyment to seeking knowledge and even socializing with other users. Although some people are using it correctly, a few people excessively use YouTube to the point where they become addicted. Addiction harms the individual. Henceforth prevention is required to make the individual not addicted or even prevent the addiction in the first place. This study aims to find the main factor that can influence someone to become addicted to YouTube and discover the factors that can help reduce the addiction to YouTube. The factor variable used to determine the cause of addiction is selected using the uses and gratification theory; Entertainment Gratification, Information Gratification, and Social Gratification. The factor variable that will help reduce addiction is; self-control, social environment, age, and virtue. The researcher took the required data sample from 428 people using the questionnaire method and analyzed the data using PLS-SEM software to determine the validity and reliability of data collected. The study shows that out of three gratifications proposed for becoming main factor for individual to use YouTube, the social gratification has the strongest influence on YouTube usage, and from the four variables given to reduce addiction, the social environment has the strongest variable to help individual on reducing addiction. |
|
Keywords: |
Addiction, Factor, Uses and Gratification, Prevention, YouTube. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
VIRTUALIZATION OF INTERNATIONAL TRADE UNDER THE CONDITIONS OF DIGITAL ECONOMY |
|
Author: |
ILONA DUMANSKA, DMYTRO VASYLKIVSKYI, IGOR ZHURBA, YAROSLAVA MUDRA, IEVGENIIA SHELEST, NATALIA BAZALIYSKA |
| Abstract: | The study is devoted to the organizational and economic principles of virtualization of international trade in a digital economy. Analytical aspects of the impact of digital economy technologies on the virtualization of international trade on the basis of systematization of digitization scenarios and expert research of intergovernmental organizations and analytical agencies are analyzed. The main preconditions and challenges for the virtualization of international trade by levels: global, international, individual state, individual trading company, taking into account the independent motivation for change and the impact of public policy. Groups of business processes of international trading companies have been identified and are suitable for the introduction of technology for their transformation, which belong to the tools of the digital economy. The methodological bases of segmentation of companies in the context of their readiness for virtualization of processes in international trade on the basis of preliminary questionnaires of respondents are substantiated. The following types of international companies are established as: non-virtualized, partially virtualized, basic virtualized, companies with outsourcing of business process virtualization and virtualized companies. It is proved that the segmentation method is effective for determining the type of international companies for compliance with the parameter “trade virtualization” and their further ranking in order to make management decisions. |
|
Keywords: |
Virtualization; Digitalization; Digital Tools; International Trade; Business Process |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
VOICE PRODUCTION FROM THE MOVEMENTS OF A HAND |
|
Author: |
RODOLFO ROMERO-HERRRERA, JESUS YALJA MONTIEL PEREZ |
| Abstract: | The system presented allows the interpretation of the movements of a hand. With sensors integrated with a Micro Bit card and the use of recorded voice, the concatenation of phonemes is generated. The result allows relating each degree of movement of a hand with the X and Y axes to reproduce a specific phoneme. Time-series analysis is performed in a range from -90o to 90o for both the X and Y axes; in such a way that by comparing the data generated and those previously stored, it is possible to relate a movement to a specific phoneme. In the concatenation of phonemes, recursion was used through graphs. Therefore, audio output is delivered. Intelligibility greater than 84% is obtained for a total of 350 phonemes. The system was developed using recursive functions in graphs. |
|
Keywords: |
BBC Micro Bit; accelerometer; degrees of hand positions; phoneme; Cartesian x, y-axes; recursive |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
USERS OPINION MINING OF TIKTOK SHOP SOCIAL MEDIA COMMERCE TO FIND BUSINESS OPPORTUNITIES FOR SMALL BUSINESSES |
|
Author: |
CHYNTIA IKA RATNAPURI, MULYANI KARMAGATRI, DIAN KURNIANINGRUM, ISTON DWIJA UTAMA, ARIS DARISMAN |
| Abstract: | Social media commerce rapidly grew during the world-changing of covid 19. The long duration of using social media provokes users to make purchases online through social media. TikTok - the most popular social media downloaded - provides their user a social media commerce experience, namely TikTokShop. The TikTok shop opened the window of opportunity for small business growth. Small business owners could maximize the potential use of TikTok shops by understanding more about TikTokShop. This research is concern to analyze the strengths and weaknesses of TikTok shop as a social media commerce platform by applying data mining. Opinion data is collected from Twitter and processed using the Naïve Bayes algorithm to test the sentiment of TikTok shop users. These positive and negative opinions are then classified and transformed with a SWOT analysis to find out the strengths and weaknesses of Tiktok Shop. This research contributes to analyzing the strengths and weaknesses of the social media commerce platform Tiktok Shop as a reference for users and platform developers. Users can see and position themselves using the platform to get optimal benefits. For platform developers, the results of this research contribute to providing insights for maximizing the features that meet the user’s expectations. |
|
Keywords: |
Data Mining, Tiktok Shop, Sentiment Analysis, Social Media Commerce |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
CLASSIFICATION OF BRAIN TUMORS: USING DEEP TRANSFER LEARNING |
|
Author: |
NOR AZURA HUSIN, MOHAMED HUSAM, MASNIDA HUSSIN |
| Abstract: | Brain tumor classification is important for diagnosing and treating cancers. Deep Learning has improved medical imaging with Artificial Intelligence (AI). Brain tumor's shape, size, and intensity make subclassification difficult. Medical imaging data is scarce. Any medical data involves privacy of the patients, hence unlike other image data, medical image data is not easily available. There are only few medical image data that is freely available for researchers. This project aims to develop a deep transfer learning model that can accurately classify brain cancers utilizing limited Medical Resonance Images (MRI) images. To achieve the goal, a modified GoogleNet model was used. Various learning algorithms were tested. The experiment also examined transfer learning and data augmentation. Finally, F1-average and confusion matrix were used to evaluate the model. Our model outperformed the state-of-the-art model in various research articles, according to performance matrices. Experimenters employed data augmentation and learning algorithms. |
|
Keywords: |
Deep Learning, Transfer Learning, Brain Tumors, Learning algorithms, Medical Imaging. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
INTEGRATED MATHEMATICAL MODELING OF AN ACCESS NETWORK AND A BACKBONE NETWORK FOR THE PHYSICAL INTERNET |
|
Author: |
KHALIL JHARNI, AZIZ AIT BASSOU, MUSTAPHA HLYAL, JAMILA EL ALAMI |
| Abstract: | As consumer needs continue to evolve, supply chains must improve and adapt by taking advantage of technological advances. Additionally, supply chains must sustainably meet those above-mentioned challenges and without further costs. In this paper, we support the Physical Internet as a framework for the development of future logistics networks. Respectively, we have addressed the design of a network for the implementation of the Physical Internet. This network allows, on the one hand, to take advantage of all optimization possibilities and on the other hand, enables the interconnection with the existing logistic networks through the π-gateways, which are an integral part of this future network. To address these characteristics, we have retained the approach that combines an access network and a backbone network without restriction of the networks' topology. Therefore, we have combined the median p-hub problem with the multi-commodity flow problem. The mathematical model we provide represents the physical Internet network as a graph. It considers and distinguishes the setup costs of the different nodes, the setup costs of the different links and their usage costs. Through the additional parameters and constraints presented, the model can adapt by changing the network configuration if necessary. The resolution of the model is not covered in this work. The main contribution of this work is that it is the first to introduce an access and backbone network approach for the design of the Physical Internet. Moreover, and unlike most contributions on this topic, it provides an integrated design of the access network and the backbone network. |
|
Keywords: |
Physical Internet, Access/Backbone networks, Logistics, Location Problem, Mixed Integer Programming |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
SENTIMENT ANALYSIS OF THE COVID-19 BOOSTER VACCINATION PROGRAM AS A REQUIREMENT FOR HOMECOMING DURING EID FITR IN INDONESIA |
|
Author: |
ANGGA PRATAMA, RAKSAKA INDRA ALHAQQ, YOVA RULDEVIYANI |
| Abstract: | The COVID-19 vaccination program was carried out to overcome the pandemic. In addition to vaccination, there is a booster vaccine program which is also an obligation for the public to be followed but has not received a good response from the public. Indonesia's government is making booster vaccination a mandatory requirement for mass homecoming in the Islamic celebration day of Eid Fitr, known as mudik Lebaran. This study aims to find out public opinion and perception regarding the booster vaccination program for mudik Lebaran using sentiment analysis. This study uses eight classification modeling: Naïve Bayes, Support Vector Machine (SVM), Decision Tree, Logistic Regression, Random Forest, K-Nearest Neighbor, AdaBoost, and XGBoost. The best classification modeling is SVM with the best accuracy score 88% and the F1 score 88%. Then this SVM model is used to predict the sentiment of 30,582 tweet data from March 22 to May 02, 2022. The results are 11,507 giving negative sentiment (37.63%) and 19,075 giving positive sentiment (62.37%). This result shows that the government's strategy in accelerating the COVID-19 booster vaccination program was well accepted by making it a requirement for mudik Lebaran. Furthers analysis with visualization of time series, shows that the sentiment had eveloved. In the first week, negative sentiment prevailed due to reactions to this policy. The Indonesian people compare the policy of the MotoGP event in Mandalika which does not require a vaccine booster. After that, in the following weeks the positive sentiment prevailed because the community realized that boosters were important to maintain their health and that of their families back home. This research shows the importance of time series visualization because sentiment can change over time. |
|
Keywords: |
COVID-19, Vaccine Booster, Mudik Lebaran, Classification, Sentiment Analysis, Time Series |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
A FINITE DIFFERENCE SCHEME FOR FRACTIONAL PARTIAL DIFFERENTIAL EQUATION |
|
Author: |
M.R. AMATTOUCH, M. HARFAOUI, A. HADADI |
| Abstract: | Fractional derivative is a new promising field of mathematics. Nowadays many researchers are interested in defining fractional derivatives that generalize the conventional derivatives. In this paper we are interested in approximating the Caputo derivative, a classical fractional derivative. The discretization of this derivative is not an easy thing and the whole works on its discretization imply heavy memory cost and large time of computations. We present in this paper a fast scheme to discretize the Caputo derivative, and apply it in solving a fractional heat equation via a finite element method. The problem statement that this work solves is the large cost of time to discretize fractional derivative by existing methods in the literacy of numerical method. Several tests cases prove the efficiency and accuracy of our proposed scheme. |
|
Keywords: |
Caputo Derivative; Fractional Heat Equation; Finite Difference Scheme. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
NO-REFERENCE QUALITY ASSESSMENT OF MEDICAL IMAGES USING CONTRAST ENHANCEMENT |
|
Author: |
GULMIRA OMAROVA, VALERY STAROVOITOV, AIDA MYRZAMURATOVA3, LAZZAT АКZULLAKYZY, ALIYA TAKUADINA, ADILBEK TANIRBERGENOV, KULAISHA BEISENBAYEVA, ZHANNA SADIRMEKOVA |
| Abstract: | Contrast distortion is often a determining factor in human perception of image quality, but little investigation has been dedicated to quality assessment of contrast-distorted images without assuming the availability of a perfect-quality reference image. In many real-world applications, images are prone to be degraded by contrast distortions during image acquisition. Quality assessment for contrast-distorted images is vital for benchmarking and optimizing the contrast-enhancement algorithms. Visual study of medical images is essential for the diagnosis of many diseases. Various contrast enhancement methods such as histogram equalization, histogram modification methods, gamma correction, etc. are used to improve the contrast of medical images. Image quality evaluation is an integral part of the contrast enhancement and image enhancement processes. Quantitative measures of digital image quality make it possible to compare the applied processing methods and choose the best of them. The article studied methods for improving the quality of x-rays. The research was carried out in several stages. Attempts were made to increase the contrast of several tens of X-ray images in order to select the best image brightness using brightness transformation methods in the MATLAB system. Contrast improvement is supported by objective scores calculated by the NIQE and BRISQUE functions that do not require reference images. As a result of successive experiments, recommendations were proposed for selecting the parameters of the gamma correction method and the adaptive histogram equalization method, where contrast enhancement is limited in order to avoid the appearance or enhancement of noise in the image. The experiment is based on the algorithms of objective non-reference quality assessment NIQE and BRISQUE. A feature of this work is the use of objective non-reference estimates to determine the quality of images. The performed experiments allow to give preference to the NIQE assessment, since it corresponded to the results of image contrast enhancement. |
|
Keywords: |
Digital X-Ray Image, Image Quality Evaluation, Image Enhancement, Contrast Enhancement |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
COMPARISON OF FOUR ML PREDICTIVE MODELS PREDICTIVE ANALYSIS OF BIG DATA |
|
Author: |
SALWA ZAKI ABD ELHADY, NEVEEN I. GHALI , AFAF ABO-ELFETOH, AMIRA M. IDREES |
| Abstract: | Big Data is the main factor in all fields of human existence be it medical, social networks, or research, it has also made inroads into education. The large size and complexity of datasets in Big Data need specialized statistical tools for analysis where python can come in handy. The Categorical component of any data set can be quantified using limited representations but evaluating it concerning the quantitative variables return a larger set of statistical inferences. This research explores the analysis of categorical and quantitative variables scalable to Big Data in education using a contemporary statistical tool. python provides multiple dimensions to statistical analysis of the dataset; this paper however explores the statistical inference rendered using the Box Plot feature through summary measures of the dataset. These statistical inferences can be used to train a Machine for predictions and classification under a certain category, one of these important analysis approaches is the Predictive analysis which is central to almost every research project. Whether the goal is to identify and describe trends and variation in populations, create new measures of key phenomena, or simply describe samples in studies aimed at identifying causal effects, predictive analyses are part of almost every empirical paper and report. many studies have shown the important directions that are extracted from predictive analysis in various sectors, and in this work, the predictive analytics applies to a big sample of educational data in Egypt census 2017 to produce estimates of the variation of educational level related to some other population features. |
|
Keywords: |
Big Data, Big Data Analysis, Machine Learning, Machine Learning Models, Predictive Analysis |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
OPTIMIZING KANBAN CONFIGURATION IN HIGH MIX, LOW VOLUME ENVIRONTMENTS BY APPLYING MULTI-OBJECTIVE GENETIC ALGORITHM |
|
Author: |
WIRAWAN ANDI NUGROHO, JAROT S. SUROSO |
| Abstract: | Kanban system is a part of the Lean Manufacturing method which aims to reduce internal logistic effort by setting materials in supermarket or inventory at the point of use in production assembly lines. Kanban is selectively chosen with relatively high runner and stable demand raw or semi-finished good materials. A High Mix, Low Volume sensor manufacturing company located in Bintan Indonesia which produces both a high variance of product portfolios and customer demand fluctuation having problems in implementing Kanban. Fluctuations in customer demands and product mix in a relatively short period of time require dynamic Kanban configuration to adapt such situations. Dynamic Kanban configuration needs to be considered to maintain the minimum cost of total Kanban inventory but yet obtain maximum total savings from the material picking process. The Optimum Kanban configuration is proposed by applying a Multi-Objective Genetic Algorithm programmed using VBA (Visual Basic Application) and run in Microsoft Excel focusing on selecting groups of materials according to total minimum Kanban inventory cost as the first objective and total maximum saving in the internal logistic effort as the second objective. The proposed model has been validated by testing using company-sourced data. The result has been compared to the minimum global optimum from data, obtaining 89% for the first objective and 93% for the second objective. |
|
Keywords: |
Kanban Configuration, Multi-Objective Genetic Algorithm, High Mix, Low Volume, Lean Manufacturing |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
ZEALOUS PARTICLE SWARM OPTIMIZATION BASED RELIABLE MULTI-LAYER PERCEPTRON NEURAL NETWORKS FOR AUTISM SPECTRUM DISORDER CLASSIFICATION |
|
Author: |
B.SURESH KUMAR ,D. JAYARAJ |
| Abstract: | Individuals with autism spectrum disorder (ASD) struggle with social interaction and learning skills throughout their lives. An early and precise diagnosis of ASD is crucial for designing an all-encompassing rehabilitation programme that enhances the individual’s quality of life and facilitates their integration into their social, familial, and professional environments. However, because of its dependence on a specialist’s opinion, the accuracy of ASD diagnoses is frequently compromised by the attendant lack of objectivity-related biases. As a result, several studies have focused on using deep learning and optimization to provide ASD early detection methods. This paper aims to provide a novel approach to ASD classification that uses neural networks in conjunction with optimization, namely Zealous Particle Swarm Optimization-based Reliable Multi-Layer Perceptron Neural Networks (ZPSO-RMLPNN). ZPSO-RMLPNN uses different protocols and fitness scores to evaluate the fMRI more deeply to confirm the presence and absence of ASD. ZPSO-RMLPNN uses threshold values to evaluate fMRI images. ZPSO-RMLPNN is evaluated with the Autism Brain Imaging Data Exchange version II (ABIDE II) dataset using standard deep learning metrics. The results make an indication that ZPSO-RMLPNN has superior performance than the current classifiers in terms of identifying ASD and non-ASD. |
|
Keywords: |
Autism, Classification, Optimization, Neural Network, Multi-Layer Perceptron, Particle Swarm |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
CONVOLUTION NEURAL NETWORK FOR BREAST CANCER DETECTION AND CLASSIFICATION – FINAL RESULTS |
|
Author: |
BASEM S. ABUNASSER, SALWANI MOHD DAUD, IHAB S. ZAQOUT, SAMY S. ABU-NASER |
| Abstract: | Objective: Early detection and precise diagnosis of breast cancer (BC) plays an essential part in enhancing the diagnosis and improving the breast cancer survival rate of patients from 30 to 50%. Through the advances of technology in healthcare, deep learning takes a significant role in handling and inspecting a great number of X-ray, MRI, CTR images. The aim of this study is to propose a deep learning model (BCCNN) to detect and classify breast cancers into eight classes: benign adenosis (BA), benign fibroadenoma (BF), benign phyllodes tumor (BPT), benign tubular adenoma (BTA), malignant ductal carcinoma (MDC), malignant lobular carcinom (MLC), malignant mucinous carcinoma (MMC), and malignant papillary carcinoma (MPC). Methods: Breast cancer MRI images were classified into BA, BF, BPT, BTA, MDC, MLC, MMC, and MPC using a proposed Deep Learning model with additional 5 fine-tuned Deep learning models consisting of Xception, InceptionV3, VGG16, MobileNet and ResNet50 trained on ImageNet database. The dataset was collected from Kaggle depository for breast cancer detection and classification. That Dataset was boosted using GAN technique. The images in the dataset have 4 magnifications (40X, 100X, 200X, 400X, and Complete Dataset). Thus we evaluated the proposed Deep Learning model and 5 pre-trained models using each dataset individually. That means we carried out a total of 30 experiments. The measurement that was used in the evaluation of all models includes: F1-score, recall, precision, accuracy. Results: The classification F1-score accuracies of Xception, InceptionV3, ResNet50, VGG16, MobileNet, and Proposed Model (BCCNN) were 97.54%, 95.33%, 98.14%, 97.67%, 93.98%, and 98.28%, respectively. Conclusion: Dataset Boosting, preprocessing and balancing played a good role in enhancing the detection and classification of breast cancer of the proposed model (BCCNN) and the fine-tuned pre-trained models accuracies greatly. The best accuracies were attained when the 400X magnification of the MRI images due to their high images resolution. |
|
Keywords: |
Breast Cancer, Deep Learning, Convolution Neural Network |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
BREAST CANCER HISTOLOGY IMAGESCLASSIFICATION BASED ON HYBRID FEATURE AND XGBOOST |
|
Author: |
HAJAR SAOUD, PR.ABDERRAHIM GHADI, PR.MOHAMED GHAILANI |
| Abstract: | Breast cancer is the most frequently diagnosed cancer in women and the most fatal ones, the numbers of incidences and mortalities increase every year.The proposition of decision support system is become a primordial need for the early diagnostic of breast cancer, many systems have been proposed in the last years for the classification of breast cancer based on histology images into its two categories benign and malignantand also into 8 sub-classes using machine learning and several methods of features extraction but nuclei segmentation is an important step that should be taken in consideration because nuclei is the key to establish a diagnosis of malignancy due to the transformation that can undergo when cancer occurs. Therefore, in this paper we will try to propose an approach for breast cancer classification in early stage based on nuclei segmentation.7909 Haemotoxylin& Eosin stained histology images were used to extract significant features for breast cancer classification into its two categories: Benign and malignant also into its eight sub-classes: adenosis, fibroadenoma, phyllodes tumor, and tubular adenoma for benign and ductal carcinoma, lobular carcinoma, mucinous carcinoma and papillary carcinoma for malignant. Image segmentation was done by Multi Otsu Thresholding algorithm followed by Watershed segmentation for images elements separation and for facilitating the morphological feature extraction. Color, texture and morphological features were extracted and classification was done by XGBoost ensemble machine learning algorithm. The proposed architecture produced an accuracy of 94.67%in 400X magnification for the binary prediction of breast cancer from histopathology images. The results are competitive compared to the results of other state-of-the-art methods. |
|
Keywords: |
Breast cancer, histology Images, Multi Otsu Thresholding, Watershed Segmentation, Machine Learning, XGBoost. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
ALGORITHM MODELING TO PREDICT STUDENTS LEARNING ACHIEVEMENT BASED ON BEHAVIORAL PARAMETERS AS THE IMPLEMENTATION OF LEARNING MANAGEMENT |
|
Author: |
AHMAD QURTUBI |
| Abstract: | The rapid development of online technology-based education has driven the implementation of intelligent campuses and provides a broad platform for information, new learning style of students and smart learning system. Supporting student success in smart learning is an issue that every university needs to solve. The paper aims to design learning achievement prediction to classify students' campus behavior characteristics, and then algorithms are used to analyze the correlation between student behavior characteristics and their academic success. This was then applied algorithms of Naive Bayes, Random Forest, K-means and C4.5. The simulation results showed that the algorithm with high predictability accuracy is Naive Bayes (71%), followed by Random Forest (63%), K-means (51%), and finally C.4.5 (39%). Then the number of criteria is selected to set the total limit and the best value that chosen by calculating the ratio of internal and outer distances. K-means method is used to analyze performance on student learning style. The best K-means algorithm has predicted the student's success well, and the average score of the academic performance. These study concluded that naïve Bayes algorithm have higher accuracy than Random Forest and C4.5 algorithms. Colleges and universities can support the student learning achievement by measuring their style initiatives for different types of students, which will not only help improve the academic performance of students but also further improve the effectiveness of the learning style of the students. The proposed model in this study could be more developed to predict student success and help them learn better. |
|
Keywords: |
Naive Bayes, Random Forest, C4.5, K-Means Clustering, Students Learning Achievement |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |
|
Title: |
MACHINE LEARNING AND DEEP LEARNING FOR FRUIT IDENTIFICATION: SYSTEMATIC REVIEW |
|
Author: |
JEISSON ENRIQUE CUEVA CARO, JORGE ISAAC NECOCHEA-CHAMORRO |
| Abstract: | Machine learning and deep learning applications are becoming increasingly popular in the agricultural industry, especially in the fruit sector, using techniques that provide the necessary advantages to transform manual practices; the objective of this study was to carry out a systematic review of the literature on Machine Learning and deep learning techniques, tools and metrics to identify fruit characteristics, using Kitchenham's methodology, which yielded 18 articles. Among the results obtained, the most used techniques, tools and metrics were: Convolutional Neural Network (CNN) and Artificial Neural Network (ANN); Python and TensorFlow, and the most used metrics to determine the effectiveness were found to be Accuracy and precision, so it can be concluded that the described techniques are considered efficient to predict certain fruit characteristics. In addition, difficulties encountered in the literature to obtain good results are mentioned. Finally, we propose some ideas for future work in the development of fruit identification. |
|
Keywords: |
Machine Learning, Identification of Fruits, Techniques, Convolutional Neural Network, Python. |
|
Source: |
Journal of Theoretical and Applied Information Technology |
| Full Text | |
|
| |