Research on Customer Churn Prediction for IoT Cloud Platforms Based on the TMRF Algorithm

Big data mining technology has deeply penetrated the field of IoT and business functions. The focus of IoT platform operators has shifted from technology development to customer-centricity. Customer churn rate has always been a core business metric for IoT cloud platform companies. Accurate customer churn prediction based on big data mining can effectively guide platform service providers in their relevant work. This paper systematically analyzes the application status of big data mining technology in customer churn prediction for IoT cloud platforms and the establishment process of customer churn prediction schemes. The scheme uses decision tree reduction and K-Means++ clustering to improve the traditional random forest algorithm, resulting in a high-precision, low-similarity TMRF (Trees Merge Random Forest) prediction algorithm. Verification of the prediction scheme using real data from a large Chinese telecommunications operator's IoT cloud service found that the TMRF algorithm significantly outperforms C4.5 and traditional random forest algorithms in terms of precision, recall, F-score, and ROC curve, demonstrating high application value.