Paulo Laque

I created an XGBoost feature importance pricing model tunned with "RandomizedSearchCV" and an NLP for Airbnb in Rio de Janeiro. The system was built in Python using scikit-learn, Natural Language Toolkit (NLTK), pandas, and NumPy.

The project involved performing feature importance analysis for Airbnb pricing in Rio de Janeiro using data preprocessing techniques, XGBoost models, "RandomizedSearchCV," and GPU acceleration for enhanced performance. Key factors impacting prices included the number of bathrooms, private rooms, specific neighborhoods, and amenities like air conditioning and smart speakers. Thorough data preparation, feature selection, and careful tuning were crucial for maximizing earnings on Airbnb.

I conducted sentiment analysis using the FinBERT model on five years of financial documents from all S&P 500 companies, specifically focusing on comments in the "management discussion and analysis" section of the 10,000 filings submitted to the Securities and Exchange Commission. This project involved developing and automating a sentiment analysis system in Python utilizing scikit-learn, transformers, Natural Language Toolkit (NLTK), pandas, and NumPy.

I used FinBERT, a deep-learning model for financial records, to generate sentiment scores for each document. I then compiled the results into a table, with each row representing an S&P 500 company and a corresponding sentiment score. I discovered that the sentiment expressed in the reports tended to be more optimistic than market sentiment, as reflected by stock prices. I noted that only 2019 showed statistical significance in predicting returns based on the sentiments expressed in the statements.

I have created a website that calculates the best credit cards available for your profile in Brazil. I developed this website using WiX and Looker (formerly Google Data Studio), which collects user input on investments, annual earnings, and expenses. It uses this information to calculate yearly fees, discounts, cashback, miles, and other rewards for each bank segment and credit card commercial rules. The cards are then ranked based on the expected profit for each one available to your profile. I own this website and earn using ads.