Machine Learning Model Based on REST API for Predicting Tenders Winner
Mardi Yudhi Putra, Rachmad Nur Hayat, Ahmad Chusyairi, Dwi Ismiyana Putri, Solikin Solikin · 2022 Seventh International Conference on Informatics and Computing (ICIC) · 2022
Machine Learning is a tool that can provide the ability to make good predictions using experience data. The problem found from the results of research observations is that the marketing department reports that PT Rekayasa Industri is currently experiencing a decline in revenue that has the potential to harm the company, one of the factors is lack of understanding and deciding whether the company should participate in project tenders or not with aspects namely price, partners, and competitors. The purpose of this study is to utilize machine learning in developing a back-end system to help the marketing department to determine the probability of winning the upcoming tender so that it can be integrated with internally owned applications. The purpose of this research is to utilize machine learning to determine the probability of winning the upcoming tender by developing a back-end system using 3 (three) classification algorithms, namely Logistic Regression, Decision Tree, and Random Forest with REST API technology. The results of this study indicate that Random Forest is the appropriate model in this case, namely predicting victory, being able to study experience data well, and of course being a model that can help minimize the analysis process carried out by the marketing department. This is evidenced by the results of the tests carried out, namely the accuracy value of 96.15%, f1 score of 97.14% and the AUC value of 0.97 (very good classification). Followed by Decision Tree with an accuracy value of 84.62%, an f1 score of 89.47% and an AUC value of only 0.78 (classification is quite good). Meanwhile, Logistic Regression resulted in less than optimal data learning performance with an accuracy value of 80.77%, an f1 score of 87.18% and the AUC value which was quite good in the classification task, which was only 0.72.