Kaggle-Bike-Sharing-Demand
In this machine learning project from kaggle, We are provided with a dataset which contains the different features such as weather, windspeed, temperature etc. as well as two columns containing the number of casual and registered and total number of bikes rented for 19 days of the month for two years and we are required to predict the total bike demand for the rest of the days of the month by analsing the historical usage patterns for the month which can help in studying various features such as mobility within the city and analysis of the data may bring out interesting underlying patterns within the data.
This repository contains 3 folders:
-
In the first approach I have implemented a decision tree.(Score obtained is 0.551)
-
In the second one I replaced the decision tree with the Random forest model.(Score obtained is 0.515) You can visit "https://www.kaggle.com/sidagar/well-commented-code-using-random-forest" to view the notebook.
-
In the third one I have used a combination of XGBoost and Random Forest.(Score obtained was 0.492) You can visit "https://www.kaggle.com/sidagar/using-random-forest-and-xgboost" to view the notebook.
PS:There are interactive and dynamic plotly charts in these files which are not rendered on the github viewer and
therefore I have uploaded the whole notebook in HTML format separately for reference which comes with all the graphs.
If you only want to see the notebook you can download and view the HTML version or visit the kaggle link to view it.
If you however wish to run the code on your own, you can either:
1. Download the ipynb format of the notebook along with the train and test sets and set the path of train test inside
the notebook and run it on your computer. Before running please make sure to install all the relevant libraries.
2. Visit kaggle links provided and fork the pinned version of the notebook and then run the notebook.
My Kaggle username is "sidagar" without quotes.