Software to Predict Computer Component Failures using Multinomial Naïve Bayes Based On User Information

Abstract

Based on data released by the International Data Corporation (IDC), the global market for personal computing devices experienced a year-over-year growth of 7.3% in the second quarter of 2024. This marks the second consecutive quarter of positive growth after nine quarters of decline. The trend suggests an increase in computer usage intensity, which is directly associated with a higher likelihood of hardware failures, particularly in components such as hard drives, RAM, and motherboards. This research aims to create a mobile-based application that can automatically classify types of computer component damage based on user-reported issues. The approach utilizes machine learning, specifically the Multinomial Naïve Bayes algorithm, combined with the Term Frequency–Inverse Document Frequency (TF-IDF) method for text feature extraction. The training data consists of categorized complaint texts according to the type of hardware problem. The resulting model was integrated into a mobile application to enable automated damage prediction. Experimental findings indicate that the proposed model performs effectively, achieving an accuracy rate of 78% in identifying computer damage categories. In summary, the developed application can help both technicians and general users diagnose potential hardware problems more accurately and efficiently. This not only speeds up the troubleshooting process but also reduces diagnostic errors and unnecessary part replacements. Moreover, the integration of Natural Language Processing (NLP) and machine learning enables the system to continuously improve its accuracy and adaptability as it learns from new data over time.