Projects

Objective:

To analyze customer sentiment for Amazon electronics using NLP techniques, uncovering key patterns in satisfaction, complaints, and review length impacts on ratings.

Tech Stack:

• Languages: Python
• Libraries: NLTK, TextBlob
• Platform: Google Colab

Key Components:

• Text Analysis: Sentiment scoring using TextBlob to evaluate polarity and word frequency analysis for common themes.
• Statistical Insights: Highlighted correlations between review length and sentiment trends.
• Scalability: Efficiently processed over 1 million reviews for actionable insights.

Conclusion:

The analysis reveals actionable insights for manufacturers and retailers, including feature improvements and marketing strategies aligned with customer feedback.

Open Paper PDF

Paper Preview:

Objective:

To predict user interest in promotional emails for targeted marketing, leveraging historical browsing and purchasing data to minimize customer annoyance and improve engagement rates.

Tech Stack:

• Languages: Python
• Libraries: Scikit-learn, Pandas, NumPy
• Techniques: Logistic Regression with L1 penalty, Pipeline Model

Key Components:

• Data Integration: Combined user demographic and activity data with browsing history for feature extraction.
• Feature Engineering: Extracted meaningful features from URL browsing data, such as visit duration and page types.
• Machine Learning Pipeline: Created a pipeline with StandardScaler for data preprocessing and Logistic Regression for classification.
• Performance Metrics: Achieved over 85% accuracy on test datasets through iterative model optimization.

Conclusion:

This project demonstrates the effectiveness of combining user activity data with machine learning techniques to predict customer engagement, providing actionable insights for targeted email marketing campaigns. The scalable pipeline allows for future integration of additional features and models.

Download Python File

Code Preview:

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.base import BaseEstimator, ClassifierMixin
import pandas as pd
import numpy as np

class UserPredictor(BaseEstimator, ClassifierMixin):
    def __init__(self):
        """Initialize the predictive model with scaling and logistic regression using L1 penalty."""
        self.model = Pipeline([
            ('scaler', StandardScaler()),
            ('log_reg', LogisticRegression(C=0.5, penalty='l1', solver='saga'))  # L1 penalty with SAGA solver
        ])

    def fit(self, X_users, X_logs, y):
        """Fit the model using user and log data. Prepare features and fit the logistic regression model."""
        data = self.prepare_features(X_users, X_logs)
        self.model.fit(data, y['clicked'])

    def predict(self, X_users, X_logs):
        """Predict whether users will click the email based on prepared features."""
        data = self.prepare_features(X_users, X_logs)
        return self.model.predict(data)
    
    def prepare_features(self, X_users, X_logs):
        """Prepare and return features for modeling, including URL-based features."""
        features = X_users[['id', 'past_purchase_amt']].copy()
        
        # Analyze URL visits
        X_logs['product_type'] = X_logs['url'].apply(lambda x: x.split('/')[-1].split('.')[0])
        url_features = pd.get_dummies(X_logs['product_type']).groupby(X_logs['id']).sum()
        
        # Time and log data features
        log_summary = X_logs.groupby('id').agg({
            'duration': ['sum', 'mean', 'std', 'min', 'max'],
            'url': 'count'
        })
        log_summary.columns = ['_'.join(col).strip() for col in log_summary.columns.values]
        log_summary.reset_index(inplace=True)
        
        data = pd.merge(features, log_summary, on='id', how='left')
        data = pd.merge(data, url_features, on='id', how='left').fillna(0)
        return data
        

Objective:

To analyze Wisconsin census data, predicting population using features like housing units and land use, with advanced geospatial and statistical methods.

Tech Stack:

• Languages: Python
• Libraries: GeoPandas, rasterio, matplotlib, seaborn
• Tools: Jupyter Notebook

Key Components:

• Data Sources: counties.geojson, tracts.shp, counties_tracts.db, and land.zip.
• Statistical Testing: CV testing (R² and explained variance score) to evaluate and compare model performance across datasets.
• Regression Models: Predicting population using housing unit and land-use data.
• Visualization: Created scatter plots, boxplots, and geographic maps to illustrate findings.
• Raster Analysis: Used masking and cell counting techniques to quantify land-use types within county boundaries, linking them to demographic features.

Conclusion:

This project provides valuable insights for urban planning, showcasing the potential of combining geospatial data with statistical models.

Download Jupyter Notebook

Code Preview:

Objective:

To analyze how different weather conditions impact traffic volume, using statistical testing to identify significant patterns.

Tech Stack:

• Languages: R
• Tools: RStudio, Google API Key

Key Components:

• Statistical Testing: Welch's T-tests to compare traffic volumes under snowy, clear, cloudy, and partly sunny conditions.
• Data Visualization: Histograms, boxplots, and interactive charts illustrating trends.
• Geographic Analysis: Mapped weather-specific traffic data using geographic roles.

Conclusion:

The analysis reveals actionable insights into how weather influences traffic patterns, informing urban planning and driver behavior studies.

Open HTML File

Project Preview:

Objective:

To visualize global COVID-19 case trends by country and time zone, providing interactive insights for comparative analysis.

Tech Stack:

• Software: Tableau
• Features: Interactive filters, geographic mapping

Key Components:

• Visualization: World map with circle sizes representing total cases and time zones color-coded for quick interpretation.
• Filters: Dropdown filters for narrowing data by specific time zones.
• Insights: Identified trends related to population density and policy responses across time zones.

Conclusion:

This dashboard showcases the potential of interactive visualization for large-scale datasets, offering valuable public health insights.

Dashboard:

Objective:

To provide an interactive platform for analyzing Minnesota Twins' player performance trends, using advanced statistical techniques to uncover insights.

Key Components:

• Regression Analysis: Predicts home runs for a given season based on player statistics like Batting Average, RBIs, Hits, and Runs.
• PCA (Principal Component Analysis): Reduces high-dimensional player performance data into key components, helping identify top performers and performance clusters.
• Interactive Visualizations: Plotly charts and tables with zoom and hover features for detailed exploration of yearly player statistics.
• Branded Interface: Styled in Minnesota Twins colors for a cohesive and professional look.

Logistic Regression Application:

An interactive platform utilizing a logistic regression model to predict the probability of patients attending their appointments.

Key Features:

• Input Panel:
  • Demographics: Dropdowns for gender and neighborhood, numeric input for age.
  • Health Conditions: Checkboxes for binary conditions such as Scholarship, Hypertension, and Alcoholism.
• Output Panel:
  • Probability Visualization: Dynamic bar chart displaying predicted attendance probabilities based on user inputs.

Analysis Application:

A visualization-focused tool analyzing seasonal, demographic, and health condition trends affecting no-show rates.

Key Features:

• Health Conditions Analysis: Compare no-show probabilities for various health conditions using logistic regression outputs.
• Seasonal Trends: Line plots highlighting month-wise no-show rates for targeted scheduling improvements.
• Demographic Trends: Grouped line plots showing no-show rates across age and gender cohorts.

Objective:

To provide Pokémon enthusiasts with an interactive tool for visualizing Pokémon stats and understanding how attributes vary by type, generation, and rarity.

Key Components:

• Customizable filters for type and generation, enabling specific or comparative analysis.
• Legendary Pokémon Comparison: Distinguish between Legendary and Non-Legendary stats.
• Interactive Visualizations: Compare Attack and Defense stats across generations and types.

Objective:

To design and build a dynamic circuit that converts sound levels into responsive LED color changes, demonstrating a hands-on approach to signal processing and embedded systems.

Personal Connection:

This project combines my interest in electronics and creative problem-solving. It allowed me to experiment with the practical applications of sensors and microcontrollers while bringing together my passion for visual feedback systems.

Key Components:

• Hardware: Arduino Uno, sound sensor, RGB LED, breadboard, resistors, and jumper wires.
• Programming: C++ code using Arduino IDE to map sound levels to RGB values.
• Signal Processing: Functions like analogRead() and analogWrite() to capture sound data and control LED outputs dynamically.
• Circuit Design: Resistors to protect the LED and solid jumper connections for seamless operation.

Objective:

To create a fun and interactive app that highlights key landmarks and entertainment options in Minneapolis and Madison while using technical frameworks to engage users.

Personal Connection:

As someone who has roots in Minneapolis and attends university in Madison, I wanted to showcase these cities’ unique offerings. This project allowed me to share my love for both places while honing my technical skills in data visualization and interactivity.

Key Components:

• Framework: Built with Python using the Dash framework for a dynamic, web-based experience.
• Map Visualizations: Integrated Plotly for interactive maps highlighting city locations.
• Dynamic Content: Included storytelling elements like a randomly generated "adventure poem" to add creativity to the technical framework.
• Deployment: Hosted on Render for accessibility and performance.

Objective:

To design and deploy a professional portfolio website that highlights my technical projects and skills in an accessible and visually appealing format.

Personal Connection:

This project represents my journey in web development, showcasing not just my technical growth but also my commitment to presenting my work in a polished and organized way. It’s a space where I can share my creativity and achievements.

Key Components:

• Languages: Built with HTML and CSS for structure and styling.
• Responsive Design: Optimized for viewing across devices with a seamless user experience.
• Dynamic Features: Includes collapsible "Read More" sections, downloadable project files, and embedded images.
• Hosting: Deployed via GitHub Pages with a custom domain (claudiaotero.com).
• Version Control: Used GitHub for iterative improvements and project tracking.