market-segmentation_purchase-prediction

Customer analysis and market segmentation based on site user demographics (Age, Gender, Salary, Purchased/Not Purchased) to identify target markets and optimize model to predict purchase behavior following ad click.

primary analysis:

• purchase models used to identify market clusters based on cluster and sub-cluster feature variants: Gender, Age, Salary, Purchased/Not Purchased
  
• K-means used to identify and label target range distributions of market clusters
  
(sub-process used to estimate distribution tiers for binary target weighted features)

prediction models:

• Logistic Regression using Standard Scalar, TTS
• Logistic Regression using Standard Scalar, TTS WoE Encoded Variables
• Decision Tree, TTS
  
  Data Tier-level exclusive 'K-wise' Optimization Models:
  *models used to determine efficacy of tier-market labeling as sub-divided within k-clusters, i.e. if distribution among sub-divided salary and age categorical tier thresholds, contributes effectively to model accuracy for classification of binary dependent variable.
  
  
• Logistic Regression using one-hot and K-folds
• Logistic regression using one-hot and TTS
• Decision Tree using one-hot and TTS
• Logistic Regression using WoE, IV and K-folds
• classification report, confusion matrix used to determine accuracy of each model at predicting Purchased/Not Purchased
  

purchase analysis
prediction models
custom tier training confluence map

jupyter notebook running pandas dataframes using matplotlib, seaborn, sklearn