Analisis Pengaruh Penerapan SMOTE pada Akurasi Random Forest dan C4.5 Dalam Klasifikasi Data Tidak Seimbang

Fanisa, 20110004 (2024) Analisis Pengaruh Penerapan SMOTE pada Akurasi Random Forest dan C4.5 Dalam Klasifikasi Data Tidak Seimbang. Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text Cover.pdf Download (1MB)
	Text Abstrak.pdf Download (49kB)
	Text Abstrack.pdf Download (197kB)
	Text Bab I.pdf Download (58kB)
	Text BAB II.pdf Download (335kB)
	Text BAB III.pdf Download (224kB)
	Text BAB IV.pdf Restricted to Registered users only Download (1MB)
	Text BAB V.pdf Download (48kB)
	Text Daftar Pustaka.pdf Download (130kB)
	Text Lampiran.pdf Restricted to Registered users only Download (58kB)

Abstract

With the rapid increase in the trend of indie game releases, developers need to increase the potential popularity of the games they release amidst a competitive environment. Previous research using logistic regression and random forest algorithms on the dataset of indie games from the steam platform using all types of features, has less than optimal performance and does not provide knowledge about what features contribute or affect the popularity of indie games. This research aims to predict the popularity of indie games on the steam platform and find out what features affect their popularity. The method used in this research is the CatBoost Algorithm which can handle categorical and numerical data simultaneously with better performance as a prediction model, as well as the Shapley Additive Explanations (SHAP) method to interpret the contribution and influence of features on the model's prediction results. The evaluation results show that the CatBoost model has an overall accuracy of 81%, with precision of 0.83, recall of 0.77, F1-score of 0.80 indicating a balanced ability of the model to identify differences between popularity classes. This is supported by the AUC value of 0.88 where the curve tends to move close to 90 degrees. The SHAP method reveals some of the most significant features that affect the results of predicting the popularity of indi games, such as the existence of steam trading cards category, RPG genre and the use of mac operating system in indi games will increase popularity. Similarly, the number of achievements provided, and the high price will also increase the popularity of indi games. However, the casual genre in indie games will reduce the popularity of indi games. Keywords : Prediction, Indie Games Popularity, CatBoost, Shapley Additive Explanations, Steam

Actions (login required)

View Item