SISTEM PREDIKSI LEAD CONVERSION MENGGUNAKAN MACHINE LEARNING MODEL PADA PT. XYZ

• Main Author: Radyan Nugroho, (Author)
• Format: Book
• Published: 2023-06-27.
• Subjects: Thesis; NonPeerReviewed
• Online Access: Link Metadata
• Notes: http://repository.upnvj.ac.id/25034/2/ABSTRAK.pdf; http://repository.upnvj.ac.id/25034/1/AWAL.pdf; http://repository.upnvj.ac.id/25034/3/BAB%201.pdf; http://repository.upnvj.ac.id/25034/4/BAB%202.pdf; http://repository.upnvj.ac.id/25034/5/BAB%203.pdf; http://repository.upnvj.ac.id/25034/6/BAB%204.pdf; http://repository.upnvj.ac.id/25034/7/BAB%205.pdf; http://repository.upnvj.ac.id/25034/8/DAFTAR%20PUSTAKA.pdf; http://repository.upnvj.ac.id/25034/10/RIWAYAT%20HIDUP.pdf; http://repository.upnvj.ac.id/25034/11/LAMPIRAN.pdf; http://repository.upnvj.ac.id/25034/9/HASIL%20PLAGIARISME.pdf; http://repository.upnvj.ac.id/25034/12/ARTIKEL%20KI.pdf
• Keywords: QA75 Electronic computers. Computer science; QA76 Computer software

The challenge over time for business practitioners is to filter potential customers from their market. However, the problem lies in the current uncertainties faced by marketing teams in market conditions. Until now, there have been many application users who have not made any transactions at all. This research is conducted to assist the marketing team in identifying the potential customer base based on the available data, specifically focusing on demographic aspects and recent activities of application users. The application of machine learning algorithms in this research aims to help the company classify potential customers using the available company data. The tree-based learning algorithms, namely decision tree, random forest, and gradient boosting, are employed as classifiers in this study. Prior to the comparison, the data undergoes modeling processes, starting from data preprocessing, and culminating in the measurement of algorithm performance using a confusion matrix. A total of 9,976 clean and analyzable data rows are obtained, which are then divided into two parts for training and testing, with an 8:2 ratio. The evaluation results of the three models reveal that the random forest algorithm achieves an accuracy score of 76%, recall of 76%, precision of 78%, F-1 score of 74%, and an AUC value of 0.836. Subsequently, the random forest algorithm is implemented into a website-based system for deployment, with the aim of assisting the marketing team in identifying characteristics of application users who have potential for conversion into customers.