It is an open project for students. After some research, my teammates and I decide to work on a house price prediction task, whose target price is either “low”, “medium”, or “high”. Beside using the basic features that are available intuitively, we spend lots of time in visualizing geographical features with given longitude and latitude. For better generate features, we also visualize many other features and decide the best way to manipulate them. After all the preprocessing procedures, we try regarding this problem as both classification (three categories) and regression problem (price prediction). At the end, XGBoosting model with additional time, text, and geographical features have the best results. Experiment results are available in the report.