AI and ML - Creating, training and a model using python in jupyter notebook
Budget: $10 – $30 USD
Jupyter Notebook Contents
The various sections in the notebook should include code, code comments and appropriate
Markup cells describing your approach chosen.
In detail, sections should include the following:
Introduction and Problem Definition
- Textual description providing an overview over the data
- A discussion on why this problem is a regression problem
- A detailed problem statement question as discussed in Lecture 3.
Data Ingestion
- Code to load the data into a suitable format to be used in the notebook
- A description of the statistical data types for each field in the file days.csv
Data Preparation
You should assume that exploratory data analysis has taken place and the following was
concluded:
- Missing values are in the ‘temp’ and ‘atemp’ columns.
- The peak usage hours are: 7-9AM and 4-7PM on working days, and 10am-4pm on
non-working days.
- At night (10pm-4am) the bike rentals are low
- If the humidity or wind-speed is high, the number of rentals decreases.
Your data preparation steps should therefore include the following:
- Fill the missing values in the temperature columns automatically with values that
would most closely mirror the actual temperature.
- Create a new field that indicates whether it is a peak time or not
- Create a new field that indicates whether it is night time or not
- Remove all fields containing information about specific dates (‘yr’, ‘mnth’, ‘dteday’),
‘casual’ and ‘registered’ and any other variables that you deem irrelevant.
- A justification (and potential application) of whether you should use data binning or
not
- Suitable encoding of the data
Data Segregation
- Code and justification for the selection and application of a suitable data split
Model Training
- Selection of two different Regression models and justification why they are suitable.
Only one of those models should be Tree-Based (e.g. Random Forest or Decision
Tree)
- Application of those models as a baseline on the data
4
- Utilisation of manual or automatic hyperparameter optimization and justification of
your choices to create “optimized” versions of each regression model
Model Evaluation
- Selection of appropriate regression metrics and a written outline why they are
suitable for this data
- A comparison of the baseline models to the “optimized” versions and an evaluation
of the results
Conclusion
- A conclusion and interpretation of the results and suggestion of potential
improvement
The various sections in the notebook should include code, code comments and appropriate
Markup cells describing your approach chosen.
In detail, sections should include the following:
Introduction and Problem Definition
- Textual description providing an overview over the data
- A discussion on why this problem is a regression problem
- A detailed problem statement question as discussed in Lecture 3.
Data Ingestion
- Code to load the data into a suitable format to be used in the notebook
- A description of the statistical data types for each field in the file days.csv
Data Preparation
You should assume that exploratory data analysis has taken place and the following was
concluded:
- Missing values are in the ‘temp’ and ‘atemp’ columns.
- The peak usage hours are: 7-9AM and 4-7PM on working days, and 10am-4pm on
non-working days.
- At night (10pm-4am) the bike rentals are low
- If the humidity or wind-speed is high, the number of rentals decreases.
Your data preparation steps should therefore include the following:
- Fill the missing values in the temperature columns automatically with values that
would most closely mirror the actual temperature.
- Create a new field that indicates whether it is a peak time or not
- Create a new field that indicates whether it is night time or not
- Remove all fields containing information about specific dates (‘yr’, ‘mnth’, ‘dteday’),
‘casual’ and ‘registered’ and any other variables that you deem irrelevant.
- A justification (and potential application) of whether you should use data binning or
not
- Suitable encoding of the data
Data Segregation
- Code and justification for the selection and application of a suitable data split
Model Training
- Selection of two different Regression models and justification why they are suitable.
Only one of those models should be Tree-Based (e.g. Random Forest or Decision
Tree)
- Application of those models as a baseline on the data
4
- Utilisation of manual or automatic hyperparameter optimization and justification of
your choices to create “optimized” versions of each regression model
Model Evaluation
- Selection of appropriate regression metrics and a written outline why they are
suitable for this data
- A comparison of the baseline models to the “optimized” versions and an evaluation
of the results
Conclusion
- A conclusion and interpretation of the results and suggestion of potential
improvement