System and method for product usage evaluation and prediction

1. A computer implemented method comprising:

receiving a list of customers using a first version of a product;

receiving customer usage and interaction data associated with the product for each customer identified on the input list of customers;

obtaining initial training data for training a machine learning evaluation model;

filtering outliers that skew data distribution from the initial training data to obtain filtered training data;

training the model with the filtered training data, wherein the model is based on a multilayer perceptron neural network and includes multiple hidden layers; and

for each customer identified on the input list of customers:

extracting evaluation features from the input customer usage and interaction data, wherein the evaluation features include at least a marketing channel used by the customer to obtain the first version of the product and data associated with user interactions with the product indicating transactions of at least the following types: linking bank accounts, adding service items, and inviting users to use the product,

processing the evaluation features by at least compensating for class imbalance and skewness,

applying the processed customer usage and interaction data evaluation features to a trained machine learning evaluation model, and

performing, by the trained machine learning evaluation model, a classification process to predict a likelihood that the customer will obtain a second version of the product, wherein the likelihood is determined according to the types of the transactions such that a degree of likelihood is increased for the customer performing transactions of each of the types relative to other customers performing transactions of fewer than all of the types.

2. The method of claim 1 , further comprising outputting a list of predicted likelihoods for each customer.

3. The method of claim 2 , wherein the output list of predicted likelihoods is prioritized based on the customers determined to be most likely to obtain the second version of the product.

4. The method of claim 1 , wherein predicting the likelihood that the customer will obtain a second version of the product comprises:

determining, using the trained machine learning evaluation model, an evaluation score for the customer;

comparing the evaluation score to a predetermined threshold;

classifying the customer as a first type when the evaluation score is less than the predetermined threshold; and

classifying the customer as a second type when the evaluation score is greater than the predetermined threshold.

5. The method of claim 4 , wherein the first type comprises a warm lead and the second type comprises a hot lead.

6. The method of claim 1 , wherein the evaluation features comprises at least one feature from user profile data.

7. The method of claim 6 , wherein the user profile data comprises an email address or contact telephone number.

8. The method of claim 1 , wherein the processing of the extracted customer usage and interaction data further comprises compensating for bad data within the extracted customer usage and interaction data.

9. A system for evaluating usage of a product, said system comprising:

a non-volatile memory; and

a processor coupled to the memory, the processor configured to:

receive a list of customers using a first version of the product;

receive customer usage and interaction data associated with the product for each customer identified on the input list of customers;

obtain initial training data for training a machine learning evaluation model;

filter outliers that skew data distribution from the initial training data to obtain filtered training data;

train the model with the filtered training data, wherein the model is based on a multilayer perceptron neural network and includes multiple hidden layers; and

for each customer identified on the input list of customers:

extracting evaluation features from the input customer usage and interaction data, wherein the evaluation features include at least a marketing channel used by the customer to obtain the first version of the product and data associated with user interactions with the product indicating transactions of at least the following types: linking bank accounts, adding service items, and inviting users to use the product,

processing the evaluation features by at least compensating for class imbalance and skewness,

applying the processed customer usage and interaction data evaluation features to a trained machine learning evaluation model, and

performing, by the trained machine learning evaluation model, a classification process to predict a likelihood that the customer will obtain a second version of the product, wherein the likelihood is determined according to the types of the transactions such that a degree of likelihood is increased for the customer performing transactions of each of the types relative to other customers performing transactions of fewer than all of the types.

10. The system of claim 9 , wherein the processor is further adapted to output a list of predicted likelihoods for each customer.

11. The system of claim 10 , wherein the output list of predicted likelihoods is prioritized based on the customers determined to be most likely to obtain the second version of the product.

12. The system of claim 9 , wherein predicting the likelihood that the customer will obtain a second version of the product comprises:

determining, using the trained machine learning evaluation model, an evaluation score for the customer;

comparing the evaluation score to a predetermined threshold;

classifying the customer as a first type when the evaluation score is less than the predetermined threshold; and

classifying the customer as a second type when the evaluation score is greater than the predetermined threshold.

13. The system of claim 12 , wherein the first type comprises a warm lead and the second type comprises a hot lead.

14. The system of claim 9 , wherein the evaluation features comprises at least one feature from user profile data.

15. The system of claim 14 , wherein the user profile data comprises an email address or contact telephone number.

16. The system of claim 9 , wherein the processing of the extracted customer usage and interaction data comprises compensating for bad data within the extracted customer usage and interaction data.