Your team has developed a tool called "SVM XXXXXX" that aims to predict the likelihood of an article being read by a credit analyst. The tool offers a range of visualizations and models to help users understand the data and make predictions.
One visualization shows the news coverage of a particular company over time, allowing users to compare the coverage of two different companies. Another visualization is a word cloud that displays the most frequently used words in the articles. The tool also includes polarity keywords to help users determine whether the words have a positive or negative sentiment.
Additionally, the tool can generate a list of the top 10 keywords with the most positive or negative sentiment. The team faced the challenge of weighing keywords when some words, such as "new," are very frequent. However, they overcame this challenge by using additional data that was not available through the virtual machine.
The tool also offers sentiment analysis of companies over time, allowing users to see how sentiment changes for a selected company. The team used several models, including Naive Bayes, SVM, Word Embedding, CNN, and BERT, to test the performance of the tool. SVM and CNN showed the best results, and the team used a confusion matrix to evaluate the accuracy of their predictions.
Finally, the team added a feature that generates text based on the most frequently used keywords in the articles. Overall, the tool offers a range of visualizations and models to help users predict the likelihood of an article being read by a credit analyst.
Log in or sign up for Devpost to join the conversation.