Diamonds Regression and Classification

This repository contains two projects on the Diamonds dataset

1. Linear Regression - predicting the diamond prices
2. Logistic Regression - classifying the prices into 4 bins

Both notebooks explore and test preproccesing, feature engineering, model training, and evaluation.

Notebooks Overview

1. Linear Regression

   • Predics diamond prices
   • Techniques: Ordinal Encoding, Log-Scaling, Feature Engineering, TrainTestSplit, and evaluation using Scatter Plots
   • Key Takeaway: log-scaling drastically improved the score: ~0.90 to ~0.97
   • Some thoughts: I had an amazing time coding this model, espacially at the time of finding out the huge difference log-scaling made while testing out scaling methods

2. Logistic Regression

   • Categorizes the prices into 4 bins
   • Techniques: Ordinal Encoding, Feature Engineering, Standard Scaling, GridSearchCV + Pipeline, and Confusion Matric for evaluation
   • Key Takeaway: Even small differences in predicted price can rezult in misclasification, overall score: ~0.92
   • Some thoughts: It is the first classifier that I ran on a regression dataset (converted into a classification dataset). Personally, the most important part of this notebook was the Confusion Matrix, showing that a score of ~0.92 can be somewhat decieving, as in reality the predictions are often very close but considered wrong because they where classified int to a bin adjecent to the correct one because of a small value difference

Usage Guide

• Python: 3.13.7
• Libraries: pandas, numpy, seaborn, scikit-learn, matplotlib

Clone repository:

git clone https://github.com/lakalex/Linear-Logistic-Regression.git

Open the notebooks in VS Code or Jupyter Notebook

Key Notes

• Both notebooks contain documented information about each model and their behavior on the dataset
• Both show the importance of understanding core model principles and data preprocessing, while showing the thought procces behind certain decisions