Estimate House Price in Tokyo: the AutoML and MLOps Approach
Managing the H2O AutoML project with Weights & Biases
Three decades after the housing bubble burst, real estate remains the cornerstone of Japanese wealth, comprising around half of household assets as of 2019. Although Japan’s population is shrinking, robust foreign investment has sustained high property prices in major cities like Tokyo. Under these market conditions, buyers frequent real estate websites like SUUMO and homes.co.jp to hunt for bargains. But with thousands of listings, finding the best deal can be overwhelming. Size, orientation, material, and last but not least, access to train stations all play a role in determining the final price.
Fortunately, a valuable resource exists to assist homebuyers. Government websites like reinfolib.mlit.go.jp offer a rich dataset, meticulously recording past real estate transactions with corresponding property details. This data provides a foundation for building a house price prediction machine learning model. Such a model is helpful as it not only provides a price estimate but also reveals how various property features impact that valuation. Buyers can use its predictions to filter the properties and make more informed decisions.
