List of files :
Used Car Price Prediction Notebook.ipynb : This is a Jupyter Noebook file which was used to generate the insights for the analysis.
Used Car Price Prediction results.pdf : This file contains final results as technical summary with insights from the analysis as pdf.
Motivation:
The evolving landscape of the personal car market, marked by a shift towards sharing services due to a daunting buying experience, inspired this project. Car salespeople, among the least trusted professionals, exacerbate the issue, particularly in the pre-owned car market. Leveraging statistical learning, this project addresses this urgency, enhancing transparency, and reshaping the industry. Here, we used a dataset from the German site of eBay containing classified advertisements for pre-owned cars.
Tech Stack:
The entire analysis was completed in Jupyter Notebook, using Python packages like NumPy, Pandas, Scikit-learn, and Matplotlib to unravel insights from the data.
Methodology:
We retrieved a dataset from Kaggle, applied data cleaning techniques, including imputation for missing values. Through exploratory data analysis and correlation matrix examination, irrelevant features were eliminated, and categorical features were converted into dummy variables. Linear and non-linear models were trained, and the best model was determined using RMSE and Accuracy metrics.
Results:
Our model, powered by the Random Forest algorithm, achieves an impressive 96% accuracy in predicting pre-owned vehicle prices. Key insights revealed customer preferences, emphasizing the significance of engine power, registration year, and vehicle type. Recommendations, detailed in the executive summary report, aim to reshape strategies in the used car market.