/Customer-Analytics-in-Python

Key: clustering, using logistic regression to build elasticity modeling for purchase probability, brand choice, and purchase quantity & deep neural network to build a black-box model to predict future customer behaviors.

Primary LanguageJupyter Notebook

Customer Analytics

Cluster analysis and dimensionality reduction to help to segment the customers.

  • We used both hierarchical and flat clustering techniques, ultimately focusing on the K-means algorithm. We also combined it with Principal Components Analysis (PCA) to reach an even better insight about our customers.
  • Customer Analytics Segmentation.ipynb
  • segmentation data.csv

Descriptive statistics and exploratory data analysis.

  • Once segmented, customers’ behavior requires some interpretation. We obtained the descriptive statistics by brand and by segment and visualizing the findings.
  • Through the descriptive analysis, we formed our hypotheses about our segments, thus ultimately setting the ground for the subsequent modeling.
  • Purchase analytics descriptive analysis
  • purchase data.csv

Elasticity modeling for purchase probability, brand choice, and purchase quantity

  • We calculated purchase probability elasticity, brand choice own price elasticity, brand choice cross-price elasticity, and purchase quantity elasticity.
  • We used linear regressions and logistic regressions.
  • Purchase analytics predictive analysis
  • purchase data.csv

Deep Learning to predict future behavior

  • Result: reached 90%+ accuracy in our predictions about the future behavior of our customers.
  • Deep Leaning Processing, Modeling and Predicting
  • Audiobooks_data.csv

Packages - NumPy, SciPy, pickle TensorFlow, and scikit-learn.