Zhiquan
My job was to train a machine learning model that can predict the connections between subjects. The project uses git for version control. -Pre-processing with the natural language data. Using regular expression to detect and deal with noise in data, formatting them and store them in database (Postgresql). -Using K-means and LDA to make clusters of the data and targets, then use decision tree to train a model that connects subjects. Both are implemented with scikit-learn. -Improving model performance by using CNN and LSTM in the NLP process to improve the quality of clusters. CNN and LSTM was implemented with keras and Tensorflow. The result was adapted by a start-up company in China and work in their recommender system (https://www.51zhiquan.com/).