Bike count prediction using BDT- Swati Kaiwar
This project is for predicting the bike count using Boosted Decision Tree Regression.
This experiment is for carrying out Machine Learning Life Cycle using Algorithm, Boosted Decision Tree Regression.
Steps:
1. Dataset:
Bike Rental UCI dataset
2. Select Columns in Dataset season,mnth,hr,holiday,weekday,workingday,weathersit,temp,atemp,hum,windspeed,cnt
Above columns are Categorical and will affect the bike rental count.
3. Edit Metadata
Select Categorical="Make Categorical"
4. Filter Based Feature Selection
Select "Chi Squared"
and select "Cnt" i.e. Number of desired features=1
5. Split Data
Splitting Mode="Split Rows", Fraction of the rows=0.7 (to send 70% of the data to the Train Model)
6. Train Model
Connect "Boosted Decision Tree Regression" to the Train Model
7. Score Model
8. Evaluate Model
After running the experiment, below are the results:
Mean Absolute Error 49.276969
Root Mean Squared Error 71.945755
Relative Absolute Error 0.345372
Relative Squared Error 0.156987
Coefficient of Determination 0.843013
Results depict that using the Boosted Decision Tree Regression is a better option than Linear Regression for this dataset because of the COD value here.