• SG AMEX CM Model: ◦ Objective: X-Sell Model, Predict customers most likely to buy another specific product via Tlemarketing Channel ◦ Data preparation: ▪ Attached the list of customer_ID(PH_ACCTID in Policy table on S6) who were contacted for X-sell with campaign outcome (Sale :1/0, Effective_Contacct:1/0) ▪ For each customer in the file extract existing policy and customer details from S6 data prior to the contact month (‘Campaign_Month’ in attached) ▪ # Policies held, product held(P/A, HI), Tenure with chubb, Premium, month since last purchase, age, gender, Channel, billing_method, billing frequency, #AMEX Policies, # non-AMEX policy, #cancelled_policies,month since last cancellation etc. ▪ AMEX Sponsor Code: Sponsor_code ('A01', 'AGH', 'AMA', 'AXB', 'AXR', 'A1A', 'AA1') ▪ Refer to S6 code that will be shared for data extraction and make the sql queries based on your understanding ◦ Data Profiling: ▪ Start with identifying key response factors and show response and customer volume distribution across each category of identified factors on a slide • Model building: ◦ Split 70:30 for train and test ◦ Run multiple ML multiple algorithms to choose best with hyperparameter tuning and validate against hold out sample ◦ Show AUC curve, feature values/contribution, decile distribution by propensity to X-Sell and other stats based on final algorithm, for In-time and out of time samples • Scoring Code: ◦ Save output as .pkl file and write scoring code to replicate the model outcome on original data