Due: 5 pmChinatime (UTC+8 Beijing) on Fri 26th May2023

Weight: 50%

Maximum scor e: 100 marks ( 100 individual marks)

Assessed lear ning outcomes:

E . Demonstr ate the ability to write code to obtain numerical solutions to mathematical pr oblems.

F . Demonstr ate the ability to display computational r esults in tabulated or gr aphical for ms.

Assessment tasks:

Overview

Employee Attrition analysis is a very typical big data prediction and analysis problem. You need to use the MATLAB programming technology you have learnt to complete the  specified data analysis and display tasks.

Submission format instructions

You need to submit two documents, ‘EmployeeAttrition.m’ , ‘Attrition_result.csv’ and ‘ID_report.pdf’ via Learning Mall Online to the correct drop box.

EmployeeAttrition.m : includes the Matlab code that can be run directly.

Attrition_result.csv: Blind test output.

ID_report.pdf : includes the Matlab code, outputs, and figures in your report.

Data introduction

You need to download the raw data set named “Attrition train.csv” and “Attrition test.csv” from LMO.

The data set includes 13 columns of house data. The meaning of data in each row is as follows:

Task 2 Main reason prediction (30 Marks)

In human resources analysis, the traditional model aims to find out the biggest reason for personnel attrition. For example, some researchers believe that salary is the main reason for attrition, while others believe that the work environment may be the main reason. Please complete the following steps to find the possible main reason for the attrition of the company.

1. Data preparation: Select 80% of the age data as the training input data in sequence and the remaining 20% as the test input data. Similarly, 80% of the attrition data should be selected as the training output data, and the remaining 20% should be chosen as the test output data. (5 marks)

2. Model training: using the training input data and training output data for model training with a linear regression algorithm. (5 marks)

3. Model test: use the test input data and test output data to verify the accuracy of the linear regression model. (5 marks)

4. Multi-dimensional analysis: construct a linear regression model and verify the accuracy of each model with “DistanceFromHome” (5 marks), and “MonthlyIncome” (5 marks).

5. Compare the model accuracy of “age”, “DistanceFromHome”, and “MonthlyIncome”, and find out the most relational element. (5 marks)

Task 3 Multi-dimensional classification (30marks)

Machine learning algorithms use a better multi-dimensional analysis method than data statistics, leading to better results. Please complete the following steps to establish a machine learning model for employee attrition and output the test results.

1. Data preparation: select 80% of the 12 dimensions of data as training input data and the remaining 20% as test input data randomly. (5 marks)

2. Model training: use the training input data with KNN algorithm to train the classification model.

(5 marks)

3. Model test: use the test input data to verify the trained classification model and check the accuracy rate. (5 marks)

4. Model tuning: traverse the K of KNN algorithm from 2-10 to find the K with the best accuracy.

(5 marks)

5. Test data load: read the data in the given “Attrition_test. csv” file. (5 marks)

6. Blind test output: use the best KNN classification model to predict the attrition of each employee and save the final results in the “Attrition_result.csv” file. (5 marks)

Sample Output

Sample of Attrition_result.csv

Marking Criteria

The following criteria will be used to assess the Coursework 2 (individual) assignment. Outstanding:

Correct output, correct variable type usage, good naming rules, good memory control, strong semantic and readability.

Appropriate:

Correct output, correct variable type usage, good naming rules, poor memory control, poor semantic and readability.

Needs improvement:

Correct output, good naming rules, wrong variable type usage,poor memory control, poor semantic and readability.

Hard to understand

Correct output, poor naming rules, wrong variable type usage,poor memory control, poor semantic and readability.

No submission or missing section

No submission or missing section including code and report