Natalie Hsu, Katherine Lam, Christina Pathrose, Ruhika Fernandes
With the cost of tuition on the rise, and the number of students taking out loans also increasing, the student debt crisis has caught the nation’s attention. However, what many people don’t realize is that this issue is also a big gender-inequality issue.
Student debt carries many consequences. Not only does it place a massive unseen burden on the smolders of many young women in the United States, it also causes a great emotional toll on these very ambitious women, deterring them from pursuing their dreams and hindering their success.
We want to empower women (hence, our name FemPow) by providing them with a means to get access to a quality education without getting tangled in the hardships associated with taking out a student loan.
What it does
We created a website that allows people, specifically women, to find the best company to get loans from. In order to do so, we will need informations about the applicant. For example, the income of the applicant, co-signer’s income, credit score of the co signer, amount the applicant plans on borrowing, whether the applicant’s credit is good or not which is used to determine how good did the applicant repay the previous loans. After getting the informations, our system will compare different companies’ loan plans and output the company that best suit the applicant’s needs.
In addition, we offer a platform in which women can communicate with other women who have been in their shoes and experienced the same concerns and feelings these new college-bound students are facing. We wish to give them a safe space in which they can voice their questions and concerns and get advice, guidance, and support from actual people who they can connect with.
How I built it
We first created a dataset with random values . We were aiming to run a linear regression model on that which would give us a clear idea as to what to have on the html css form that we were going to take as user input .
We obtained a bad linear regression model as the values were too randomly generated and hence had to discard the model as we obtained a very low root mean square value.
We then conducted a research found out that most women were deprived of getting federal
loans approved and hence took private loans instead . We then built our linear regression model to observe how it influenced the interest rate keeping , ApplicantIncome,Co-applicantIncome,credit_score_cosigner,LoanAmount,Recent.payment.history for co signer in mind and found that root mean square value generated by the model was high as compared to the p value and hence decided this model was relevant and hence would have all these inputs on the UI. Next based on the data set we had to determine which of these lenders exhibited similar interest rates and used a hierarchical complete linkage clustering algorithm in order to obtain the similarities between companies based on the interest rate Post completion we made used of a classification algorithm of decision trees in order to obtain the range of interest rate based on credit_score_cosigner,Recent.payment.history and realized that most of the applicants whose co - signers had a recent payment history marked as “N” were subjected to higher interest rates having the median as 7.7 whereas applicants having cosigners having a good payment history end up with an median interest rate of 4.65. When the user made use of the user interface on entering the variables that fit the linear regression model the interest rate could be given to them based on the classification algorithm and then the associated lender company would be recommended to them based on the clustering algorithm.
Challenges I ran into
We first ran into challenge with finding good data that we can used in our program for comparing loan options. We wanted to find data that specifically demonstrates the difference between women and men when borrowing loans and repaying loans. However, there are inadequate amount of data that contains the informations that we want. As a solution to the problem, we created a predictive model that recommends the best interest rate option to the user.
The second challenge we ran into was that there was linear between our modeled data. It was mainly caused because we randomly generated the data. In order to solve this problem, we added the option for user to include their credit when searching for the best loan option. By doing so, we created a better representative data based on the applicant’s credit.
Accomplishments that I'm proud of
We are proud that we were able to overcome the challenges that we faced with the help of the mentor and reached our goal. Even though our final product is slightly different than what we originally expected it to be, we are still proud that we were able to accomplish so much at a short period of time, considering that we were not very sure what we wanted to do at the beginning. In addition to that, we are proud that we are able to create this website that will surely raise awareness for one of the many issues causes by gender inequality.
What I learned
We learned that sometimes we would have to change our original plan according to the current situation, and that not everything would work accordingly. Two of our teammates starting to learn more about HTML. We also learned that we could decide the input that we want. We were made aware that the HTML css user interface could be connected to R using Rook so we could use it to do a real time prediction based on data received through the form and provide a good interest rate based on the form input.
What's next for FEMPOW
We would send out survey to women who actually get loans in order to get a better data set for a better database. We would also set up an email account or social media for people to subscribe to, where we would post inspiring stories of women who had to get loan in college but were able pay it off and become successful. We would also update any features we added to our website or services.