NextCarPrediction
Download as PPTX, PDF
1 like63 views
1) 19% of existing customers become repeat customers, purchasing a second or third car from the same dealership. 2) The document analyzes purchase history data to determine which subsequent car models repeat customers are most likely to purchase after their initial car. 3) Several predictive models are proposed, including decision trees, to more accurately predict a repeat customer's next vehicle based on additional customer profile data like age, income, gender, and occupation. Better predicting customer preferences could help improve marketing strategies.
NextCarPrediction
- 1. Predicting Repeat Customers Next Car 3st March 2016 V1.0
- 2. Repeat Customer Proportion Type of Customer No. of Cars Non repeat Customer 468,558 Customer who purchased 2 cars 177,447 Customer who purchased 3 cars 75,965 468,558, 81% 88,723, 15% 25,321, 4% No. of Customers, % Non repeat customer Bought 2nd Car Bought 3rd Car 19% of existing customers become repeat customers!
- 3. Count Analysis Method:- S.N. Customer ID No. of Contracts MM-YYYY Product 1 1000102580 3 03-2003 CAMRY 2 1000102580 3 07-2005 LANDCRUISER WAGON 3 1000102580 3 03-2008 FORTUNER Previous Car Subsequent Car CAMRY LANDCRUISER WAGON CAMRY FORTUNER LANDCRUISER WAGON FORTUNER From History of Sales Observing Preference of Buying Subsequent Car Counting number of purchase of same first car and same subsequent car for all repeat customers. PreviousSubsequent FORTUNER LANDCRUISER WAGEN CAMRY 1 1 LANDCRUISER WAGEN 1
- 4. How first car owners preferred their next car(s)? First Car CAMRY PRADO LANDCRUISER WAGON COROLLA FORTUNER Yaris Sedan YARIS (H/B) 16% 2% 1% 28% 8% 17% COROLLA 22% 2% 2% 32% 6% 12% CAMRY 31% 3% 1% 20% 6% 8% FORTUNER 13% 7% 2% 21% 16% 14% Aurion 20% 2% 2% 20% 3% 10% INNOVA 13% 2% 1% 22% 10% 14% Subsequent Car * Indicating selected products due to space constraint Only Active products used for subsequent car preference for marketing relevance
- 5. Top 3 prediction with distribution:- 0% 5% 10% 15% 20% 25% 30% COROLLA CAMRY Yaris Sedan COROLLA
- 6. Top 5 prediction with distribution:- 0% 5% 10% 15% 20% 25% 30% 35% CAMRY COROLLA Yaris Sedan HILUX DOUBLE CAB(IMV3) FORTUNER CAMRY
- 7. Top 3 prediction with distribution:- 0% 5% 10% 15% 20% 25% 30% COROLLA Yaris Sedan CAMRY YARIS (H/B)
- 8. Top 3 prediction with distribution:- 0% 5% 10% 15% 20% 25% 30% COROLLA Yaris Sedan CAMRY YARIS (H/B)
- 9. How can we improve this knowledge? Decision Tree Model: Non-linear Prediction Model for multi- category classification through hierarchical segmentation of the data, by partitioning recursively. A tree structure of rules over the input variables are used to classify or predict according to the target variable. Decisions upon additional internal information we have about the repeat customers can be used as additional input variables. Buying preference generally depends upon multifactor ( Additional Data we have of our Customers) Marital Status Nationality Work Place Job Guest Age Range Guest Income Range Guest Gender Government/ Non Government
- 10. Data Preparation 1. Exploring the significance of variables 2. Data Partition:- 55% Training, 45% Validation. Chisquaredplotofvariablesbytheir chi-squaredStatistics. Variableworthplot:Worthofvariablesby theirworthinpredictingthetargetvariable.
- 11. Best to see in E-Miner Window
- 12. Model Interpretation: Misclassification Rate Lower Misclassification rate represents a good model. Model performance is good as indicated by the training and validation model results are close.
- 13. Model Interpretation: Average Square Error Lower error demonstrates minimal target selection bias by the model.
- 14. Thank you!