Machine learning Problem 1. Regression.
Budget: $10 – $100 AUD
The data in Data.zip contains socio-economic data from the 1990 US census for various US communities, and the number of violent crimes per capita (in the column ViolentCrimesPerPop). The purpose
of the data is to explore the link between the various socio-economic factors and crime. The provided data has been split into training (communities train), validation (communities val), and testing (communities test) sets.
Your Task: Given the provided data, you are to:
1. Train a linear regression model to predict the number of violent crimes per captia from the socio-economic data.
2. Train a Ridge regression model to predict the number of violent crimes per captia from the socio-economic data.
3. Train a LASSO regression model to predict the number of violent crimes per captia from the socio-economic data.
For your analysis, you are to use all provided data (i.e. DO NOT perform any filtering or
selection to remove columns and/or rows). For LASSO and Ridge models, the validation
dataset should be used to select the optimal value of λ. All models should be evaluated
on the test set. Evaluations should consider the predictive power of the model, the model
complexity, and the model validity.
Your response must include sections that address the following:
Discussion and justification of any pre-processing performed, such as standardisation.
Details of the three trained models, including key parameters such as values for λ for
the LASSO and Ridge models, and a brief discussion of how these were selected. (
An evaluation comparing the three models. Your evaluation should consider model
accuracy and model validity. Discussion of accuracy/validity should also consider the
socio-economic nature of this data, and if or how this impacts the the required accuracy
and/or utility of the resultant models. You may also wish to discuss the relevance of
terms within the model. The use of tables to summarise results, and figures to highlight
details such as model validity, is strongly recommended.
Socio-economic modelling is incredibly complex and often dealing with sensitive data
and variables. Briefly discuss the ethical concerns that need to be considered when
evaluating the models developed for this problem. This discussion may include ref-
erence to potential confounding variables, or to limitations in the modelling schemes
used.
of the data is to explore the link between the various socio-economic factors and crime. The provided data has been split into training (communities train), validation (communities val), and testing (communities test) sets.
Your Task: Given the provided data, you are to:
1. Train a linear regression model to predict the number of violent crimes per captia from the socio-economic data.
2. Train a Ridge regression model to predict the number of violent crimes per captia from the socio-economic data.
3. Train a LASSO regression model to predict the number of violent crimes per captia from the socio-economic data.
For your analysis, you are to use all provided data (i.e. DO NOT perform any filtering or
selection to remove columns and/or rows). For LASSO and Ridge models, the validation
dataset should be used to select the optimal value of λ. All models should be evaluated
on the test set. Evaluations should consider the predictive power of the model, the model
complexity, and the model validity.
Your response must include sections that address the following:
Discussion and justification of any pre-processing performed, such as standardisation.
Details of the three trained models, including key parameters such as values for λ for
the LASSO and Ridge models, and a brief discussion of how these were selected. (
An evaluation comparing the three models. Your evaluation should consider model
accuracy and model validity. Discussion of accuracy/validity should also consider the
socio-economic nature of this data, and if or how this impacts the the required accuracy
and/or utility of the resultant models. You may also wish to discuss the relevance of
terms within the model. The use of tables to summarise results, and figures to highlight
details such as model validity, is strongly recommended.
Socio-economic modelling is incredibly complex and often dealing with sensitive data
and variables. Briefly discuss the ethical concerns that need to be considered when
evaluating the models developed for this problem. This discussion may include ref-
erence to potential confounding variables, or to limitations in the modelling schemes
used.