IJATIS: Indonesian Journal of Applied Technology and Innovation Science

Comparison of Support Vector Machine, Random Forest, and C4.5 Algorithms for Customer Loss Prediction

Bima Maulana — 2025-02-28

Loss of customers has been discussed and many studies have been conducted, starting from using the Bayesian network algorithm, Decision tree, random vorest, Support vector machine, and neyral network Algorithms Support Vector Machine (SVM), Random Forest, and Decision Tree or C4.5 are algorithms used for prediction and have several advantages Random forest has the advantage of being able to combine many predictions from decision trees that have a tendency to reduce overfitting. This research uses the C4.5 algorithm, SVM and random forest. Research shows that the Random Forest method has the highest accuracy of 87.02% compared to the Support Vector Machine and Decision Tree methods. In contrast, Decision Tree gets low accuracy results with a value of 78.52%. Experimental results show that the Random forest method for customer loss prediction achieves an average classification accuracy of 4% - 9% higher than the Support Vector Machine and Decision Tree methods.

Implementation of Supervised Learning Algorithm on Spotify Music Genre Classification

Muhammad Fiqri — 2025-02-28

Spotify is a music streaming application that has been around since 2008. In the application, users can compile a playlist of songs they want to listen to. Users can determine the name of the singer, type of music, music genre and tempo of the music they want to listen to play as needed. The genre received by each user from his device will produce different recommendations, this is due to the classification process based on music listening behavior, such as songs that are often, rarely, or even never listened to or played at all by users. Therefore, the process of classifying music genres on spotify with the help of machine learning using supervised learning algorithms with algorithms namely Naïve Bayes, K-Nearest Neighbors (K-NN), Random Forest and Decision Tree with the aim of comparing the accuracy of each algorithm so as to get the best model for calcification. The results of this study obtained Random Forest has the highest accuracy value of 79.40%, followed by Decision Tree at 79.30%. In the next position Naïve Bayes with an accuracy value of 77.28%, the algorithm with the lowest accuracy is K-NN with an accuracy value of 60.74%. Meanwhile, evaluation with the t-test algorithm with the best performance is obtained from the Random Forest algorithm with a value of 0.794. It can be concluded that the best algorithm in music genre classification on Spotify is using Random Forest.

Determining Zakat Recipients Using Simple Multi Attribute Rating Technique with Analytic Hierarchy Process Eigen Preference

Muhammad Ridho Anugrah — 2025-02-28

Paying zakat for Muslims is an obligation to alleviate the burden of recipients. However, difficulties arise in determining the right individuals for zakat distribution because each type of mustahiq or zakat recipient can seem similar to one and another therefore become hard distinguish. This research aims to enhance accuracy using a Decision Support System (DSS) with criteria like Number of Dependents, Income, Occupation, Home Ownership, Marital Status, House Walls, House Floors, and House Roof. The Analytic Hierarchy Process (AHP) method simplifies unstructured problems into a hierarchy, and the Simple Multi-Attribute Rating Technique (SMART) offers flexibility in analysis. Decision outcomes are rankings with the highest scores, ordering those most deserving of zakat. Weighting results highlight Number of Dependents with the highest weight at 0.335 for determining zakat recipients. Based on ranking, alternative A1 secures the top position with a score of 0.077.

Analyzing Customer Sentiment Towards Marketplace Reviews Using Classification Algorithms

Nabiilah Nabiilah — 2025-02-28

Numerous online marketplaces like Shopee and Lazada have been developed in Indonesia due to the rapid growth of e-commerce. The Shopee and Lazada apps link buyers and sellers in transactions to purchase and sell products and services. About 100 million users have downloaded both applications as of this writing. Since releasing these programs, the community has voiced various thoughts and complaints. Based on this, user sentiment regarding the Shopee and Lazada applications on the Google Play Store is determined using sentiment analysis using the K-Nearest Neighbor (KNN), Nave Bayes, and Support Vector Machine (SVM) algorithms. Data selection, pre-processing, transformation, data mining, and assessment are the five stages of the Knowledge Discovery in Databases (KDD) approach. For each E-commerce application, 2000 reviews were used as the data. With an accuracy of 85.71% for Gaussian-NB modeling for the Lazada dataset and an accuracy of 85.67% for Bernoulli-NB modeling for the Shopee dataset, the Naive Bayes algorithm has the highest accuracy in experiments on each dataset.

Comparative Analysis of the Combination of AHP-SAW and AHP-WP in Making Decisions on Hiring New Employees

Rizki Andreas — 2025-02-28

This paper's web-based employee recruitment the goal is to help Human Resource Development (HRD) managers automatically calculate criterion weights and alternative weights, refinement of potential employees and a faster selection process. Recommendation system applications use Combination of Simple Additive Weighting (SAW) and Analytic Hierarchy Process (AHP). The AHP method determines the importance of each professional criterion is at the moment. SAW, on the other hand, determines the position or priority of a potential employee, calculated from alternative options. In the AHP method, criteria influence the outcome of a decision. The resulting calculations are examined using the specified priority weights to see which criteria are most important. The weight value for the CI criterion was 0.0603, and the CR value was 0.0538. However, a sensitivity analysis of criterion priorities is required to examine the extent to which small effects on criterion weights change the ranking of alternatives. Based on the ranking results using AHP-WP, Fajar ranked first with a preference value of 0.1037289. You can also see how important the selection criteria are to the ranking results.

Comparison of Supervised Learning Algorithms for Predicting Airline Passenger Satisfaction

Agil Irman Fadri — 2025-03-03

Service quality and airline passenger satisfaction are the main factors in business success in the modern aviation industry. This research compares the performance of supervised learning algorithms, namely K-Nearest Neighbor (K-NN), Naïve Bayes, Decision Tree, Random Forest, and Support Vector Machine (SVM), to predict passenger satisfaction. The k-fold cross-validation method with k=20 was applied to ensure comprehensive model evaluation by dividing the data proportionally. Using a high value of ???? was chosen to optimize the stability of the model estimates, reduce the risk of overfitting, and produce more accurate evaluation metrics. The research results show that the Random Forest algorithm provides the highest accuracy of 95.78%, followed by Decision Tree (93.82%) and K-NN (91.85%). These results indicate that the Random Forest algorithm better classifies passenger satisfaction than other algorithms. This research confirms the potential of machine learning algorithms as a practical solution in data analysis to support strategic decision-making, especially for airlines that want to improve customer experience.

Obesity Prediction Using Machine Learning Algorithms

Hanifatus Syahidah — 2025-03-03

This study aims to develop a prediction model for obesity levels by utilizing five machine learning algorithms, namely K-Nearest Neighbors (K-NN), Naïve Bayes Classifier (NBC), Decision Tree, Random Forest, and Support Vector Machine (SVM). The data used in this study were obtained from Kaggle, consisting of 2111 data with 17 attributes covering lifestyle and demographic factors. The research process involved data collection, pre-processing, data division using the Holdout Split method (70% training data and 30% testing data), and the application of machine learning algorithms. Performance evaluation used accuracy, precision, recall, and F1 score metrics. The results showed that the Random Forest algorithm had the best performance with an accuracy of 92.29%, followed by Decision Tree at 90.54%, K-NN at 83.44%, and NBC and SVM which reached 59.15% and 59.08%, respectively. Confusion matrix analysis revealed that NBC and SVM had difficulty distinguishing certain obesity classes. Based on these findings, it can be concluded that Random Forest is the most effective algorithm in predicting obesity levels. The results of this study are expected to contribute to developing a more accurate obesity prediction system that can be implemented in the real world.