Generating customer’s credit behavior with deep generative models

Banks collect data x 1 in loan applications to decide whether grant credit and accepted generate new 2 throughout the period. Hence, banks have two measurement-modalities, which provide a complete picture about customers. If we can conditioned on keeping relationship between these modalities, behavior scoring may be enabled simultaneously (at time is obtained) support cross-selling, launching of products or marketing campaigns. Therefore, develop novel conditional bi-modal discriminative (CBMD) model for scoring, able based classify outcome loans an unified framework. The idea behind CBMD learn joint (among modalities) latent representations that are useful using available during application process. classifier introduced encourages generative process accurately. Further, optimizes objective function this research, maximizes mutual information representation z modality improve model. We benchmark our proposed outperforms other multi-learning models. Similarly, classification performance tested under different scenarios it achieves higher par compared state-of-the-art multi-modal learning