Good day everyone. I’m currently working on an insurance recommendation project. For both train and test, each row corresponds to a customer, assigned a unique customer ID (‘ID’). There is some information on the customer (when they joined, birth year etc). The customer’s occupation (‘occupation_code’) and occupation category (‘occupation_category_code’) are also provided, along with the branch code of the office they visit. The final 21 columns correspond to the 21 products on offer.
In Train, there is a 1 in the relevant column for each product that a customer has. Test is similar, except that for each customer ONE product has been removed (the 1 replaced with a 0). my goal is then to build a model to predict the missing product. I’ve tried xgboostClassifier but the logloss used like loss function is not enough small (0.12 in test). Please some advice or help on which approach or model would be appropriate to tackle this. Please find below the link of my work: https://jovian.ai/ngnie-christian/insurance-recommendation-for-zinmat-company