Python_Supervised_Learning

This repo is still being developed. Some small projects are done as an example but mostly as a reference for fundamentals with a heavy emphasis on data cleaning, exploration, statistical analysis and modeling. As time permits, more models will be added and complexity will grow within the various projects.

Coming Soon: Section on hyperparameter tuning!

Linear Regression

• Simple Linear Regression Guide for Machine Learning

  • A simple breakdown of the model using one independent and one dependent variable, the theory and metrics used. More to come on the theory.

• Bike Demand Predictions using MLM

  • Coming Soon!
  • Full feature analysis will include statistical procedures, testing for assumtions, and MLR modeling using a large real world data set.

• Housing Prices (Psuedo Data)

  • Purpose of this example is to layout a simplified approach to Multiple LM modeling.

• eCommerce LM Mini Project (Psuedo Data)

  • This Multiple Linear Regression model explores the variables associated with webiste sessions compared to mobile sessions to predict annual spending amounts from customers.

Logistic Regression - To be completed sometime this year (credit default LR)! This is a rather extensive project. Last update on 2/21/2022

• Credit Default: A Logistic Regression Analysis of customer credit card/default behavior.

  • Data Cleaning

  • Model Evaluation

• Titanic Survival Predictions Using Logistic Regression

  • This model is a foundational example of exploring the data, cleaning, training, predicting and evaluating metrics.

• Adversting Predcitions Using Logistic Regression

  • A foundational example of building a logistic regression model used to predict if a customer clicked on an advertisment based on selected features.

K Nearest Neigbors

• Anonymous Classification

  • This example provides the base algorithm use case and an optimization for error rate using the elbow method for K.

• Anonymous Classification Project

  • Same as the example classification above but with a larger and different anonymous data set. The concepts covered are menat to strengthen the fundamental task of using and optimizing the model.

Decision Trees & Random Forrests

• Kyphosis Decision Tree

  • This example uses a decision tree to analyze patients that had undergone an operation for a spinal condition called Kyphosis. The decision tree is used to predict whether or not a patient was healed several months post op.

• Lending Tree Decision Tree/RF Analysis (real data)

  • This analysis completes an exploratory analysis, tests a decision tree and compares the results to a random forest using the same data. All data is real data from lending tree circa 2010ish.