Analisis Pengaruh Durasi Live Selling Produk Fashion terhadap Total Penjualan Menggunakan Decision Tree

The growth of live selling on e-commerce platforms has encouraged sellers to understand the factors influencing sales performance. However, many sellers are still unaware of which variables most significantly affect total sales during live sessions. This study aims to develop a web-based visualization system called RekomLive to analyze the influence of broadcast duration, number of viewers, and interactions on total fashion product sales using the Decision Tree algorithm. The dataset consists of 122 data collected through direct observation of live selling sessions on the Shopee platform. Numerical data were transformed using the Equal Width Binning method into three categories, namely low, medium, and high, then split at a 70:30 ratio into 85 training data and 37 testing data. The results show that the viewers variable is the most influential factor, with the highest information gain value of 0.1288 and a contribution of 84.3%, followed by interaction at 15.7%, while duration does not have a significant influence. The model achieved an accuracy of 89.19%, precision of 91.67%, and recall of 97.06%. The RekomLive system successfully presents the analysis results visually and recommends a minimum broadcast duration of 240 minutes to maximize high sales opportunities. These findings can be used as an evaluation basis for sellers in developing live selling strategies for fashion products.