Information technology predicting the sale prices of houses in king-county by machine learning methods

Abstract

The sale and purchase of the real estate, in particular housing, and houses are extremely important for our life. Most people turn to real estate agencies to realtors in order to purchase quality housing and at the same time at the best price for the buyer. You should rely not only on personal assessment or assessment of third-party experts but also use price prediction systems that, using the features of the house (area, number of floors, location, number of bedrooms, year of construction, etc.), are able to predict its possible price. The report is devoted to the task of improving the accuracy of predicting the sale price of houses in King-County using machine learning methods by creating an information technology for predicting this price. The analysis of sales and purchases of real estate has been carried out, and signs that have an impact on the pricing of houses have been previously proposed. The dataset was selected, and its main features were described, preliminary data cleaning was carried out, exploratory data analysis was carried out, a rule for filtering anomalous values for the selected dataset was proposed, many possible models were selected, they were trained, and the optimal one was selected, the result of the models was presented and analyzed, for comparison the prediction accuracy of similar solutions is given. An optimal LGBM regression model was obtained, and its application made it possible to obtain a prediction accuracy of 0.876, which is more than 0.82, as in the best analog. The scientific novelty lies in the fact that the information technology for predicting the sale price of houses in King-County has been further developed using machine learning methods, which makes it possible to increase the accuracy of such a prediction compared to analogs.