Description
Background:
In this assignment, we will continue to work on the “2020 Kaggle Machine Learning & Data Science
Survey” dataset.
The purpose of this challenge was to “tell a data story about a subset of the data science community
represented in this survey, through a combination of both narrative text and data exploration.” More
information on the competition, data, and prizes can be found on: https://www.kaggle.com/c/kaggle-survey2020/data
The dataset provided (Kaggle_survey_2020_responses.csv) contains the survey results provided by Kaggle.
The survey results from 20036 participants are shown in 355 columns, representing survey questions. Not
all questions are answered by each participant, and responses contain various data types. In the dataset,
column Q24 “What is your current yearly compensation (approximate $USD)?” contains the ordinary
categorical target variable. The original data (kaggle_survey_2020_responses.csv) has been transformed
to clean_kaggle_data_2020.csv as per the code given in KaggleSalary_DataSet.ipynb. In the dataset to
be used for Assignment 2 (clean_kaggle_data_2020.csv- File to be read in notebook for this Assignment,
You should work with the clean dataset for this assignment), rows with the null values of salaries have
been dropped. In addition, two columns (‘Q24_Encoded’ and ‘Q24_buckets’) has been added at the end.
Column ‘Q24_buckets’ (Target Variable for Assignment 2) have been obtained by combining some salary
buckets in the column ‘Q24’. Column ‘Q24_Encoded’ has been obtained by label encoding the column
‘Q24_buckets’.
The purpose of this assignment is to train, validate, and tune multi-class ordinary classification models that
can classify, given a set of survey responses by a data scientist, what a survey respondent’s current yearly
compensation bucket is.
2/5 MIE 1624 Introduction to Data Science and Analytics – Assignment 2
Classification is a supervised machine learning approach used to assign a discrete value of one variable
when given the values of others. Many types of machine learning models can be used for training
classification problems, such as logistic regression, decision trees, kNN, SVM, random forest, gradientboosted decision trees and neural networks. In this assignment you are required to use the ordinal logistic
regression algorithm, but feel free to experiment with other algorithms.
For the purposes of this assignment, any subset of data can be used for data exploration and for classification
purposes. For example, you may focus only on one country, exclude features, or engineer new features. If
a subset of data is chosen, it must contain at least 5000 training examples. You must justify and explain
why you are selecting a subset of the data, and how it may affect the model.
Data is often split into training and testing data. The training data is typically further divided to create
validation sets, either by just splitting, if enough data exists, or by using cross-validation within the training
set. The model can be iteratively improved by tuning the hyperparameters or by feature selection.
Learning objectives:
1. Understand how to clean and prepare data for machine learning, including working with multiple
data types, incomplete data, and categorical data. Perform data standardization/normalization, if
necessary, prior to modeling.
2. Understand how to apply machine learning algorithms (logistic regression) to the task of
classification.
3. Improve on skills and competencies required to compare performance of classification algorithms,
including application of performance measurements, and visualization of comparisons.
4. Understand how to improve the performance of your model.
5. Improve on skill and competencies required to collate and present domain specific, evidence-based
insights.
Questions:
The following sections should be included but the order does not need to be followed. The discussion for
each section is included in that section’s marks.
1. [2pts] Data cleaning:
While the data is made ready for analysis, several values are missing, and some features are
categorical. Note that some values that appear “null” indicate that a survey respondent did not select
that given option from a multiple-choice list. For example – “Which of the following hosted
notebook products do you use on a regular basis? (Select all that apply) – Selected Choice – Binder
/ JupyterHub”.
For the data cleaning step, handle missing values however you see fit and justify your approach.
Provide some insight on why you think the values are missing and how your approach might impact
3/5 MIE 1624 Introduction to Data Science and Analytics – Assignment 2
the overall analysis. Suggestions include filling the missing values with a certain value (e.g. mode
for categorical data) and completely removing the features with missing values. Secondly, convert
categorical data into numerical data by encoding and explain why you used this particular encoding
method.
These tasks can be done interchangeably, e.g., encoding can be done first.
2. [3pts] Exploratory data analysis and feature selection:
For the exploratory data analysis step, visualize the order of feature importance. Some possible
methods include correlation plot, or a similar method. Given the data, which of the original
attributes in the data are most related to a survey respondent’s yearly compensation?
Explain how feature engineering is a useful tool in machine learning in the context of the tasks in
this assignment. Apply feature engineering and then select the features to be used for analysis
either manually or through some feature selection algorithm (e.g. regularized regression).
Not all features need to be used; features can be removed or added as desired. If the resulting
number of features is very high, dimensionality reduction can also be used (e.g. PCA). Use at
least one feature selection technique – describe the technique and provide justification on why
you selected that set of features.
3. [4pts] Model implementation:
Implement ordinal logistic regression algorithm on the training data using 10-fold crossvalidation. How does your model accuracy compare across the folds? What is average and variance
of accuracy for folds? Treating each value of hyperparameter(s) as a new model, which model
performed best? Give the reason based on bias-variance trade-off. An output of your algorithm
should be a probability of belonging to each of the salary buckets. Apply scaling/normalization of
features, if necessary, and justify the reason why scaling/normalization is (not) needed.
4. [3pts] Model tuning:
Identify all hyperparameters in your model. Select two hyperparameters for model tuning and
justify your selection. Improve the performance of the models from the previous step with
hyperparameter tuning and select a final optimal model using grid search based on a metric (or
metrics) that you choose. Choosing an optimal model for a given task (comparing multiple
classifiers on a specific domain) requires selecting performance measures, for example accuracy,
precision, recall and/or F1-score to compare the model performance. Justify the metric you
selected. There is no minimum model accuracy, as long as your methodology is reasonable and
well explained.
4/5 MIE 1624 Introduction to Data Science and Analytics – Assignment 2
5. [4pts] Testing & Discussion:
Use your optimal model to make classifications on the test set. How does your model perform on
the test set vs. the training set? The overall fit of the model, how to increase the accuracy (test,
training)? Is it overfitting or underfitting? Why? Plot the distribution of true target variable values
and their predictions on both the training set and test set. What insight have you gained from the
dataset and your trained classification model?
Insufficient discussion will lead to the deduction of marks.
Submission:
1) Produce an IPython Notebook (.ipynb file) detailing the analysis you performed to answer the questions
for the given data set, save or print it to a .pdf file.
2) Produce a 3-page report explaining your response to each question for the given data set and detailing
the analysis you performed. When writing the report, make sure to explain for each step, what you are
doing, why it is important, and the pros and cons of that approach.
Tools:
● Software:
○ Python Version 3.X is required for this assignment. our code should run on the
CognitiveClass Virtual Lab https://labs.cognitiveclass.ai (Kernel 3). All libraries are
allowed but here is a list of the major libraries you might consider: Numpy, Scipy, Sklearn,
Matplotlib, Pandas.
○ No other tool or software besides Python and its component libraries can be used to touch
the data files. For instance, using Microsoft Excel to clean the data is not allowed.
○ Read the required data file from the same directory as your notebook on the CognitiveClass
Virtual Lab – for example, pd.read_csv(“clean_kaggle_data_2020.csv”).
● Required data files:
○ clean_kaggle_data_2020.csv: file to be read in notebook for this Assignment
○ The data file cannot be altered by any means. The notebook will be run using the local
version of this data file. Do not save anything to file within the notebook and read it back.
● Auxiliary files:
○ kaggle_survey_2020_responses.csv: original survey responses.
○ kaggle_survey_2020_answer_choices.pdf: the questions and answer choices in the
survey.
○ kaggle_survey_2020_methodology.pdf: the methodology and survey flow logic of the
survey.
○ KaggleSalary_DataSet.ipynb: the code used to transform the original survey responses
(kaggle_survey_2020_responses.csv) to the clean dataset clean_kaggle_data_2020.csv
5/5 MIE 1624 Introduction to Data Science and Analytics – Assignment 2
What to submit:
1. Submit via Quercus a Jupyter (IPython) notebook containing your implementation and motivation
for all the steps of the analysis and a printed or saved pdf with the following naming convention:
lastname_studentnumber_assignment2_notebook.ipynb
lastname_studentnumber_assignment2_notebook.pdf
Make sure that you comment your code appropriately and describe each step in sufficient detail.
Respect the above convention when naming your file, making sure that all letters are lowercase and
underscores are used as shown. A program that cannot be evaluated because it varies from
specifications will receive zero marks.
2. Submit a report in PDF (up to 3 pages) along with the source files (.docx/.doc/.tex) including the
findings from your analysis. Use the following naming conventions
lastname_studentnumber_assignment2_report.pdf
lastname_studentnumber_assignment2_report ( .doc /.docx / .tex )
For latex, it is acceptable to submit a compressed file as source file.
Late submissions will receive a standard penalty:
● up to one hour late – no penalty
● one day late – 15% penalty
● two days late – 30% penalty
● three days late – 45% penalty
● more than three days late – 0 mark
Other requirements and tips:
1. A large portion of marks are allocated to analysis and justification. Full marks will not be given for
code alone.
2. Output must be shown and readable in the notebook. The only files that can be read into the
notebook are the files posted in the assignment without modification. All work must be done within
the notebook.
3. Ensure the code runs in full before submitting. Open the code in CognitiveClass Virtual Lab
(Kernel 3) and navigate to Kernel -> Restart Kernel and Run all Cells. Ensure that there are no
errors.
4. Do not re-run cross-validation (it can run for a very long time). When cross-validation is finished,
output (print) the results (optimal model parameters). Hard-code the results in the model parameters
and comment out the cross-validation code used to generate the optimal parameters.
5. You have a lot of freedom with however you want to approach each step and with whatever library
or function you want to use. As open-ended as the problem seems, the emphasis of the assignment
is for you to be able to explain the reasoning behind every step.
6. The output of the classifier when evaluated on the training set must be the same as the output of
the classifier when evaluated on the testing set, but you may clean and prepare the data as you see
fit for the training set and the testing set.
7. When evaluating the performance of your algorithm, keep in mind that there can be an inherent
trade-off between the results on various performance measures.