Starbucks Capstone Challenge

1. Dependencies

• pandas
• numpy
• math
• json
• seaborn
• boto3
• os
• sagemaker
• matplotlib
• dateutil
• sklearn
• plotly

2. Notebooks

• 1.Starbucks_Capstone_notebook-Data_Preprocessing.ipynb

  • Contains all the preprocessing steps, feature engineering and upload preprocessed training and test data to Amazon S3.

• 2.Starbucks_Capstone_notebook-Training_an_Estimator.ipynb

  • Contains the training, evaluation, deployment and result steps.

3. Project description

The goal of this project consists in using Amazon SageMaker’s supervised learning model, DeepAR, to predict how much someone will spend based in the next days. Knowing that information, it is possible to discover which demographic groups respond best to which offer type and create a recommendation system that suggests customized offers. According to DeepAR forecasting algorithm documentation, it is a method for finding time-based patterns and forecasting scalar (one-dimensional) time series using recurrent neural networks (RNN).

The machine learning model which was used to predict how much a user will spend next day will be deployed using Amazon SageMaker. To be consumed, the model will use a web-app that will allow a user pass his ID, using it as a trigger to Amazon API Gateway which will trigger a lambda function to preprocess the input data and send preprocessed input data to the model. The main tasks involved to create the recommendation system are the following:

1. Exploring and cleaning the data;
2. Choosing a model;
3. Training an estimator that can predict how much someone will spent according to the offer received;
4. Building an API using Amazon API Gateway and make the estimator run on a web-app;
5. Make the web-app predicts a how much a user will spend in each offer to the next day.

This was the final project of the Machine Learning Engineer Nanodegree program by Udacity.