Embedded Files
Statement of Authenticity
I, Natalia Fajardo, confirm that all the work presented in this portfolio has been completed by me. The projects and information included are authentic representations of my skills, goals, and accomplishments. Any collaborators mentioned are aware of this publication and have agreed to have their contributions shared as part of my team within this portfolio.
Personal Loan Prediction Model
Personal Loan Prediction Model
In this project, I implemented different ensembles and decision trees in Python to predict whether to approve or deny a personal loan application. To address the class imbalance, I used SMOTE to oversample, since the original data was 90.4% classified as denied and 9.6% as approved.
The code walks you through each step: data partition, oversampling (SMOTE), and various model fitting to identify the most accurate one. Click the button to view the project on GitHub.
Results
Most models—such as Decision Tree, Bagging, AdaBoost, and Random Forest—perform well in predicting denials but have room for improvement in identifying approvals. On the other hand, the top-performing models are Gradient Boosting, with an accuracy of 97.2%, a precision of 99% for denials, and 85% for approvals, and Random Forest with hyperparameter tuning, which achieved the highest accuracy at 97.3%. The latter performs similarly to Gradient Boosting, with a slight improvement of 0.1% in overall accuracy; therefore, the Random Forest with hyperparameter tuning is the best model overall.
Predicting Students at Risk of Depression in Post-Secondary Institutions
Predicting Students at Risk of Depression in Post-Secondary Institutions
With SAS® Enterprise Miner™, we predicted student depression risk through Logistic Regression, Decision Trees, and Neural Networks. We also identified key risk factors and proposed intervention strategies. The project was presented at the 2025 SAS Curiosity Cup Competition.
CuriosityCup_MindPulseAnalytics.pdfInventory Management Dashboard
Inventory Management Dashboard
This dashboard built in Tableau helps determine whether inventory is being managed effectively.
It combines sales and inventory data to identify overstocked and understocked products, track how fast inventory moves, and show how many units are in stock. The goal is to support smarter decisions, reduce waste, avoid stockouts, and keep operations running smoothly.
Prediction of Car Prices with Neural Networks
Prediction of Car Prices with Neural Networks
In this project, I implemented Neural Networks to predict car prices using scikit-learn.
The code walks through each step: data processing, scaling, partitioning, and model fitting using Grid Search to find the best combination of hyperparameters. Click the button to view the project in Google Colab.
Let’s collaborate!
Let’s collaborate!
Page updated
Google Sites
Report abuse