Purchase Experience Score prediction
This page contains an example of using the MLSToolbox Code Generator tool for graphically defining an ML pipeline and generating the corresponding Python code. The defined ML pipeline aims to train a linear regression model for predicting the users' purchase experience score in the largest department store in Brazilian marketplaces.
The selected datasets for this example and their structure may be found in the Brazilian E-Commerce Public Dataset by Olist. In total, the datasets have 52 columns. The datasets have been merged following the data schema provided in Brazilian E-Commerce Public Dataset by Olist to obtain a single dataset.
The pipeline is composed of the following stages:
-
Data Collection: gets the data of the customers, purchases and review comments about purchase experiences from the olis_data_dataset dataset -
Data Cleaning: replace null values for the median for 4 columns and replace null values for a specific text "No Review" for the review_comment_message column -
Feature Engineering: selects the columns to train the model and splits the columns into the features used to train and the truth or target values. -
Model Training: splits the features and truth into data to train the model and into data for testing the model. Moreover it trains the resulting model using a linear regressor algorithm. -
Model Evaluation: calculates the accuracy, MSE, RMSE and R2 metrics of the trained model.
All the information of the purchase experience score prediction example and the generated code of the ML pipeline can be found on the pipeline examples repository.