Oscar Calzada

Our machine learning algorithm takes in a URL corresponding to a JPEG image. It processes that URL image and converts it into an array of vectors (e.g. numbers). At this point, we employed transfer learning to use a pre-trained convolutional neural network (NN) 50 layers deep called, ResNet-50. This NN,... [Read More]

Using natural language processing (NLP) and machine learning techniques, we conducted sentiment analysis on Yelp User’s text reviews to develop a web application that assigns an “adjusted” ratings score (1-5) to free text. The training data we used to train our model was for all 100k reviews of tacos. [Read More]

An application to pull and analyze the content of user tweets and then predict which user is more like to say a given new tweet. I developed this app by leveraging both Twitter and Basilica’s API. Employing transfer learning, I used Basilica to transform or “embed” high-dimensional text data into... [Read More]

Using historical data (1880-2016) for monthly average global temperature anomalies in degrees Celsius, I developed a model to predict temperature anomalies into the future, which represents global temperature rise. [Read More]