I finished my Original Work!

After working on my original work for more than 3 weeks, I finally completed my original work. I was able to create 4 projects in total: Digit Classification, Sentiment Analysis on Movie Reviews, Text Generation to create a Play, and a Customer Churn Analysis.

My Digits Classification project used a neural network and trained on 60,000 (28 * 28 pixel) images and tested the algorithm on another set of 10,000 images to achieve an accuracy of 97.8%. Furthermore, the images are then converted into one-dimensional vectors which enables the neural network to make use of the data and apply gradient descent and backpropagation to successfully predict based on its past errors.

My second and third projects used recurrent neural networks to predict the sentiment (positive or negative) on movie reviews and to generate text (character by character) to create a play. I was able to use a recurrent neural network and a Long Short Term Memory (LSTM) recurrent neural network to use the previous word sequence as context for the next word sequence and create an accurate prediction.

My last project used the KNIME Analytics Platform to create different clusters of customers for Customer Churn Analysis (for a telecommunication business). It used a K-means algorithm to create 3 different centroids of customers to help predict which customers are likely to leave the telecommunication service. It used many measures such as call minutes, number of calls, account length, customer service calls, etc. to successfully organize the clusters.

So see you after winter break, same place, same time!