Ahmad Ferdi Dwi Nugroho, Hasbi Firmansyah, Wahyu Asriyani

Penerapan Data Mining untuk Prediksi Evaluasi Mobil dengan Metode Decision Tree

Introduction

Penerapan data mining untuk prediksi evaluasi mobil dengan metode decision tree. Prediksi evaluasi mobil menggunakan Data Mining dan metode Decision Tree C4.5. Penelitian ini mencapai akurasi 92.94%, mengidentifikasi faktor Safety sebagai penentu utama.

Abstract

The evaluation of motor vehicle acceptability is a multicriteria classification problem involving a complex set of qualitative attributes. This research aims to apply Data Mining techniques using the C4.5 Decision Tree Algorithm to predict the car evaluation outcome (Car Evaluation Dataset), categorized into unacceptable, acceptable, good, and very good classes. The classification model was constructed based on six categorical input attributes and implemented using RapidMiner Studio software with a 10-Fold Cross Validation scheme. The main objectives of this study were to measure the model's accuracy and to identify the priority sequence of attributes most influential in determining the car evaluation class. The test results show that the C4.5 model achieved an accuracy level of 92.94%. Furthermore, the model identified the Safety attribute as the most dominant factor affecting the evaluation outcome, followed by Persont These findings validate the effectiveness of the Decision Tree Algorithm in providing a predictive and interpretive solution for complex multicriteria scoring systems.

Review

This paper presents a clear application of data mining techniques to a practical, multicriteria classification problem: predicting motor vehicle acceptability. The authors aim to leverage the C4.5 Decision Tree Algorithm to categorize cars into four predefined classes (unacceptable, acceptable, good, very good) based on a set of qualitative attributes. The choice of the C4.5 algorithm is appropriate for such a task, given its capability to handle categorical data and produce interpretable models, which is particularly valuable when dealing with complex evaluation criteria. The problem itself is highly relevant, reflecting the intricate decision-making processes involved in assessing complex products. The methodology employed appears robust, utilizing the widely recognized Car Evaluation Dataset and implementing the classification model in RapidMiner Studio. The use of a 10-Fold Cross Validation scheme significantly strengthens the reliability of the reported performance metrics, mitigating the risk of overfitting and ensuring the generalizability of the model. A key finding is the high accuracy level achieved by the C4.5 model, at 92.94%, which is commendable for a complex classification task. Beyond mere prediction, the study's ability to identify the priority sequence of influential attributes, pinpointing "Safety" as the most dominant factor followed by "Persont," adds substantial interpretive value, a significant advantage of decision tree models. In conclusion, this research successfully validates the effectiveness of the C4.5 Decision Tree Algorithm in providing both predictive and interpretive solutions for complex multicriteria scoring systems. The high accuracy, combined with the clear identification of the most influential attributes, offers valuable insights that could inform future car evaluation frameworks or product design considerations. This study contributes positively to the application of data mining in real-world scenarios where understanding feature importance is as crucial as achieving high predictive performance.

Full Text

Comments

You need to be logged in to post a comment.