Predicting Home Prices in Ames, IA Using Regression Models
The objective of the project was to find a model minimizing the mean root squared error between the log of predicted housing sale prices in our model and actual housing models. The log of the sale prices was used as the metric in order to minimize the effect of extremely expensive and cheap sale prices on our model.
We took the steps of cleaning our data set by taking care of missing values. We attempted to impute missing values using mode, mean, classification, and a number of other methods. In order to do this we needed to organize our data into numeric, categorical, and ordinal types. From there, we started our feature engineering by creating dummy values for the categorical variables. We then explored the effects of PCA analysis, lasso regression, and decision trees to find importance of certain field. Our studies are detailed in this presentation.
Feel free to review our project contents in our Github repository.