FiyinfobaO/used_cars_price_prediction

An end-to-end project on predicting the prices of used cars in Nigeria # used_cars_price_prediction This is an end-to-end Data science project on predicting the prices of used cars in Nigeria. The project was carried out in 5 key stages: 1. Data scraping from the cars 45 site. 2. Data cleaning and preprocessing. 3. Exploratory data analysis(EDA) on the cleaned data. 4. Model building using various regression algorithms. 5. Model deployment in Heroku using Flask. # Data Scraping The data was scraped using the scraper notebook to get 4,183 cars with each car having the following information: * The car manufacturer/make * The car model * The year the car was made * The mileage(km) * The car price * The car grade * The type of fuel used * The selling condition of the car * The transmission of the car * The car color # Data Cleaning and Preprocessing The following steps were taken in order to clean the data using this notebook: * The data was checked for null values where it was discovered that all the columns apart from manufacturer and price had null values present. * The null values was handled on a column by column basis; dropping the null values for some and filling the null values manually based on further analysis and by checking the cars45 site. * Some columns were not in their appropriate data type so this was corrected. * Further preprocessing was done on the *color* column to replace the ambiguous color names to simpler names. # Exploratory data analysis(EDA) By carrying out EDA on the cleaned data as seen in this notebook, various patterns were discovered in the dependent and independent variables. For example: By plotting the distribution of the car prices and mileage values separately using distplot, it was seen that both the price and mileage distribution was positive skewed with most of the datapoints lieing on the left side so the price data had more of lower price values and the mileage had more cars of lower mileage kilometer values values. The presence of outliers was also discovered using boxplot in the price and m …