Residential Property Price Forecasting Model for Central Pangasinan

Abstract

This quantitative-experimental study aims to create a price forecasting model for the residential real properties of the fourteen municipalities and cities of central Pangasinan using supervised learning classification algorithms (linear regression and decision tree) to predict whether the real property’s value will increase or decrease in the future and classic statistical forecasting techniques (straight line, moving average, simple linear regression, and multiple linear regression) to predict the rate of increase or decrease using a -+5% margin of error. Data used were derived from the Residential Real Estate Price Index (RREPI) of the Banko Sentral ng Pilipinas (BSP) from 2016 to 2021, Zonal Valuations (ZV) from the Bureau of Internal Revenue (BIR) from 1990 to 2023, and the Housing Cost Construction Index (HCCI) from the Philippine Statistics Authority (PSA) from 2006 to 2021 following an 80:20 training-testing data split ratio. The resulting model utilizing the RandomForest algorithm shows a significant 93% accuracy rate and 93% precision rate. Results further showed that machine learning-based algorithms performed better than the four classic statistical forecasting techniques evaluated with multiple linear regression obtaining the lowest average prediction distance point of 12.46% as compared to Random Forest which achieved 4.32%.