Turning Data Into Business Value

Intelligent loan approval system for P2P lending platform. Current solution relies solely on human review of forms, making the process time-consuming, difficult to scale, and prone to subjective decision-making.

Historical loan application records combined with detailed web log data capturing user interactions with the online application form.

ML classifier for loan approval was developed based on submitted form content, behavioral signals and enriched demographic data. Thorough data analysis was also conducted to reveal insights about applicant segments and loan types.

Predict whether potential clients will default in payments for a leading mobile operator. Existing analytical model was inaccurate and incapable to provide risk score for new clients.

ERP, CRM and CDR (Call Detail Records) data. For existing clients financial and descriptive data was available through business modules, only CDR and limited demographic data was available for new clients.

Combination of traditional classification algorithms with graph based features. The model utilized CDRs to provide default prediction for new clients.