Apartment Rental Price Prediction

Project Overview

This project aims to build a machine learning model to predict apartment rental prices in Moscow. Our goal is to improve the Mean Absolute Percentage Error (MAPE) from 50% to 30% or lower.

Objectives

1. Understanding the Data

• We analyzed the dataset (Exploratory Data Analysis - EDA) to find patterns and trends.
• Identified key factors affecting rental prices.

2. Cleaning the Data

• Removed missing values and duplicates to ensure data quality.
• Standardized data formats for better consistency.

3. Feature Engineering

• Created new features (e.g., Building Type, Floor Number) to improve predictions.
• Transformed categorical variables and scaled numerical values.

4. Preparing Data for Model Training

• Finalized the dataset with relevant features.
• Ensured all data was structured correctly for machine learning algorithms.

Key Findings

• Apartment size, number of rooms, distance to city center and metro station strongly influence rental prices.
• Cleaning the dataset improved the reliability of the data.
• Adding extra features improved the model’s performance.

Visualizations

• Graphs showing how price changes with different features.
• Correlation matrix displaying relationships between variables.