OPIM 3511: Final Project Instructions
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
OPIM 3511: Final Project Instructions
General Instructions:
• Use dataset that is not used in examples/HW in our class.
• You can perform classification or regression - your choice.
Things to submit:
• Project Report: 6~8 pages, double spaces, including references.
• Dataset: a copy of all your datasets in .csv.
• Code: An .ipynb notebook with headers and comments – please make it easy for me to grade.
• Upload all the files to HuskyCT as a .zip file.
Detailed Descriptions:
|
Item |
Description |
Total Points |
|
Data Description & Project objective |
Background of how dataset is collected, what is your objective. |
5 |
|
Literature Review |
Give a background on the problem you are trying to solve. Use ten quality sources from blogs, research articles, websites, news stories. List your ten references at the end of the notebook. |
5 |
|
Exploratory Data Analysis |
Create three interesting plots and two interesting tables to include as part of your report. These must be created in Python. Perform data imputation if necessary. (Impute with mean or median will be okay) |
2 * 5 = 10 |
|
Modeling |
Use an 80/20 split on the data. Set the random seed equal to one of your student IDs (7 digit code) when splitting. Only use numeric variables. Examine the column names and data types (via df.info()) and delete any duplicate columns or categorical columns. Fit two models. For example, if you are doing regression, you can fit your data with multiple linear regression and decision tree regressor. NO simple linear regression. If you find your model is overfitted on the training data (i.e., R2=1 or accuracy=1), do hyperparameter tuning or ask help from me. |
10 |
|
Modal Evaluation |
Regression: scatterplots and R2, MAE, MSE Classification: confusion matrix and accuracy report |
5 |
|
Analysis & Discussion |
I want you to really dig into the results: How did the error metrics change across your different experiments/models? Is one ML model ‘better’ than another? … Also, relate your results to other studies on the same dataset. Reference at least three other different studies and compare your algorithm and results. Your model does not need to beat others. |
10 |
|
Conclusion |
As part of your final report, make sure you have at least five to ten sentences describing what you did in the report, and what you might do differently next time if you had more data/time/etc. |
5 |
|
Total |
|
50 |
2022-03-09