hypotheses.tex

\documentclass{report}
\usepackage[utf8]{inputenc}

\title{Financial Complaints Project Future Hypotheses}
\author{Anuj Singla}
\date{September 29th, 2022}

\setcounter{secnumdepth}{3}

\begin{document}

\maketitle

\chapter{Current Project}

\section{Optimizing the ML Model}
Only one scikit-learn Logistic Regression model was used and was biased toward classifying a complaint in the "non-dispute" category.
\subsection{If an ensemble method such as XGBoost is optimized for the data set, then the model will make less biased predictions than the Logistic Regression model.}
\subsection{If the data is balanced to include less non-disputes, then the ML classifier will train with less bias.}
\subsection{If stronger feature analysis and engineering is performed on the data set, then the ML model will have greater performance with respect to consumer disputes.}

\section{Data Insights}
The following features were compared to consumer disputes: product, state, submission method, and issue.
\subsection{If a product is more commonly used within a financial institution, then it will have more complaints.}
\subsection{If a consumer is located in a state on the western coast of the United States, then they are more likely to dispute their resolution.}
\subsection{If a consumer uses an online method of submitting their complaint, then they are more likely to dispute their resolution.}
\subsection{If a consumer submits an issue for servicing payments, then they are more likely to dispute their resolution.}

\section{Additional Dispute Prediction Project Implementations}
\subsection{If sentiment analysis on issues is used as a feature in the ML Model, then the classifier will have stronger performance.}
\subsection{If dimensionality reduction with principal component analysis (PCA) is used, then strong, new visualizations and insights can be drawn from the data.}


\chapter{Additional Ideas}

\section{Data Exploration}
\subsection{Interactive map that highlights different zip codes and their attributes.}
\subsection{Advanced sentiment analysis on product, sub-product, issue, and sub-issue.}

\section{ML Prediction}
\subsection{Predicting how the company will resolve the complaint}

\end{document}