House Prices | Machine Learning project

Predict sales prices using advanced regression techniques

2. Model Testing for Numeric Features

Baseline score and RandomizedSearchCV searching for best params using:
. DecisionTreeRegressor, AdaBoostReegressor,
  GradientBoostingRegressor, XGBRegressor
. RandomForestRegressor (extracting feature importances)
. and stacking.

. Load libraries
. Load datasets (upload datasets on GCP storage)
. Create scoring metric (rmsle)

. Preprocessing: encoding (ordinal and nominal categorical
  features, numerical features, cyclical features),
  full preprocessing pipeline;
. Feature Investigation: numerical features (VIF to measure
  multicolinearity), ordinal features (spearman method),
  univariate feature selection (SelectPercentile)

. Feature overview (numerical vs categorical features)
. Preprocessing (impute, scale, encode)
. Estimator (with DecisionTreeRegressor)
. Cross validation
. Prediction

. Best score using StackingRegressor (on test set locally):
  0.12119
. Actual score valuated on Kaggle (top 12% on leaderboard):
  0.12396

. Data loading and target engineering
. Apply preprocessing pipeline
. Create RMSE score
. Linear models (LinearRegression, Ridge)
. KNeighborsRegressor, SVR,
. Trees (DecisionTreeRegressor),
. Random Forest (RandomForestRegressor),
. Boosteed Trees (AdaBoostRegressor,
GradientBoostingRegressor),
. Stacking (VotingRegressor, StackingRegressor),
. XGBRegressor.