Merge remote-tracking branch 'refs/remotes/origin/main'

midterm.md

@@ -4,13 +4,46 @@ layout: home
 ---
 
 # Introduction / Background
+Banks and financial institutions act as money lenders and provide loans, as well as credit lines for millions of customers worldwide. However, for them to operate in a profitable manner and reduce the risk of their customers defaulting on payments for loans, they have to be able to accurately assess an individual's potential to default on their credit (in general, their overall credibility). This is also highly beneficial for them to determine the amount of interest that they should be charging on a given loan, and overall, whether they should offer the loan in the first place. This project is based around building a binary classification model that can be used to predict whether a borrower will default on their loan (meaning that they are unable to pay back a loan provided by a financial institution). A review of relevant literature showed us that privately securitized adjustable-rate mortgages have much more variance in performance when compared to retained mortgages, underscoring the importance of providing financial institutions with the best tools to aid decision making processes [(Krainer & Laderman, 2013)](https://doi.org/10.1007/s10693-013-0161-7‌). Delving into the literature revolving around this topic also greatly helped influence decisions made for this report, which will be elaborated on further in later sections.
+
+For this project, our team opted to use the dataset provided by the Home Credit Group from a Kaggle competition. This data set has over 100 features about loan applicants and their credit risk. The dataset can be found at [this link](https://www.kaggle.com/c/home-credit-default-risk/data). In later sections, we discuss how we processed, cleaned, and reduced the data set to create a binary classification model to determine if applicants will default or not.
 
 # Problem Definition
+In today's volatile housing market, numerous potential borrowers face challenges in securing loans, often hindered by limited or absent credit histories. This situation breeds opportunities for unscrupulous lenders to exploit such individuals, while traditional banks miss out on viable clients due to irregular assessment methods. Emphasizing the crucial need for accurate predictions of loan repayment, it's imperative to ensure that qualified borrowers aren't unjustly denied. Equally vital is the early identification of likely defaulters to avert financial losses. Relying solely on manual risk assessment is not only time-consuming but also subject to human biases and inconsistency. Therefore, the development of a model to categorize individuals based on their repayment risk is essential. Such a model can help financial institutions streamline the evaluation process, enhance consistency in risk assessment, and significantly lower the incidence of loan defaults.
 
 # Data Collection
 
 # Methods
 
 # Results and Discussion 
 
-# Contribution Table
+# Team Contributions
+
+## Contribution Table
+
+| Yash Gupta | Reetesh Sudhakar | Nityam Bhachawat | Mark Nathaniel Glinberg |
+| ---------- | ---------------- | ---------------- | ----------------------- |
+| Methods (Classifier, Data Pre-processing), Potential Results & Discussion (Classifier Metrics) | Project Website & Documentation, Introduction & Background, Problem Definition, Literature Review, Methods (Data Pre-processing) | Dataset Exploration & Pre-processing (manual feature selection selection via code), Classifier code, Results (Metric Analysis) | Project Timeline, Project Introduction & Background, Literature Review, Dataset Visualization and Exploration (manual feature selection & processing), Data Pre-processing |
+
+## Project Timeline and Gantt Chart
+
+Our team used the Gantt Chart that was created earlier in the semester to track our progress and maintain accountability between one another. We checked off certain boxes as our work progressed, and the updated Gantt Chart is included below. To access view the Excel file and download it, please [click here](resources/GanttChart.xlsx "download").
+
+![Midterm Report - Gantt Chart](resources/midterm-gantt-chart.png)
+
+Moving forward, our team has established the following goals and timelines for the final portion of the project:
+- Explore the dataset further (1 - 2 weeks): Though the dataset that we worked with up until this point had over 100 features, it was a small portion of the provided dataset. For the final portion of the project, we want to integrate features from other portions of the dataset. 
+- Create custom features (0 - 1 week): From examining the dataset, we noticed that certain features can be combined in some aspect to create a new feature that may be more telling (e.g. a ratio, a multiplicative factor, etc.). Creating these features will help us have a better grasp of the data, also allowing us to feed in a better cleaned dataset into our classification models. 
+- Improve Data Processing Methods: Something that our group did to ensure that there were no missing values was to use the Simple Imputer model from the Scikit-Learn library. Filling values with a constant could lead to distortions of the data, which could affect model predictions. In the future, we hope to use a more sophisticated method to take underlying patterns in the data into account.
+- Improve Classification Models (1 week): Based on the performance metrics that were computed up until this point, we hope to improve upon those metrics as we process more robust, telling data. We also hope to explore different configurations of binary classifiers and how they can better predict home credit defaulting.
+
+To assess these changes and improvements, we will continually analyze the performance metrics of our classifier, and we will also continue to create visualizations of our data as we clean it and process more features. 
+
+---
+
+# References
+
+1. Bao, W., Lianju, N., & Yue, K. (2019). Integration of unsupervised and supervised machine learning algorithms for credit risk assessment. Expert Systems with Applications, 128, 301–315. https://doi.org/10.1016/j.eswa.2019.02.033
+2. de Castro Vieira, J. R., Barboza, F., Sobreiro, V. A., & Kimura, H. (2019). Machine learning models for credit analysis improvements: Predicting low-income families’ default. Applied Soft Computing, 83, 105640. https://doi.org/10.1016/j.asoc.2019.105640
+3. Emad Azhar Ali, S., Sajjad Hussain Rizvi, S., Lai, F.-W., Faizan Ali, R., & Ali Jan, A. (2021). Predicting Delinquency on Mortgage Loans: An Exhaustive Parametric Comparison of Machine Learning Techniques. Vol12 - Issue 1, Volume 12(Issue 1), 1–13. https://doi.org/10.24867/ijiem-2021-1-272
+4. Krainer, J., & Laderman, E. (2013). Mortgage Loan Securitization and Relative Loan Performance. Journal of Financial Services Research, 45(1), 39–66. https://doi.org/10.1007/s10693-013-0161-7‌