AutoOne Price Prediction

AutoOne is a fictional company which buys and sells used cars in more than 30 countries across Europe.

The heart of their business is to chose prices for buying and selling cars while making a profit.

They have lots of information on car sales and they want to be able to make data driven decisions when deciding what are the best prices.

Therefore, in this exercise we will use the data available to build a statistical model for predicting prices for a given car.

We will address data cleaning, exploratory data analysis, linear regression and machine learning algorithms.

Contents:

• Data cleaning
  • Getting the data
  • Missing data
  • Exploring and transforming the features
• Exploratory data analysis - features - TO DO
• Exploratory data analysis - price
  • Analysis of sample distribution
  • Discussion about outliers
  • Log transformation on prices
  • Balancing the data
  • Conclusion
• Linear regression - OLS Model
  • Train-test split
  • Feature Selection
  • Fitting the model
  • More feature selection: p-values
  • Model Scoring & Evaluation
  • Residual analysis
• Linear regression - sklearn
  • Feature selection
  • Normalization and encoding of features
  • Hyper-parameter tuning and Cross Validation
  • Linear regression model
  • Regularized models - Ridge regularization
  • Regularized models - Lasso
  • Huber Regressor
  • Comparing models - MSE and \(R^2\)
  • Most important features
  • Comparing the models - conclusions
• Modeling with machine learning
  • Hyper-parameter tuning and Cross Validation
  • Decision Tree Regressor
  • k-Nearest Neighbors
  • Random Forests
  • Gradient tree boosting
  • AdaBoost
  • Comparing models - MSE and \(R^2\)
  • Conclusions
• Python scripts
• About me