List view
By March the 31st the system for weekly updates from twitter should be set up together with clustering of the gathered data. A large portion of the group about 5 people will at first work on this part. Deliverables: set of keywords for the clustering, database of twitter users in India, optimal querying for posts related to basic commodities, identification and crawling/querying of other relevant sites, Clustering of tweets and other online content using Spark's MLlib. A team of 4 people will work on researching the most suited machine learning techniques, e.g. "Recurrent Neural Networks" or "Support Vector Regression" and Time Series Analysis techniques that allow sequence prediction as well as incorporation of the additional indicators from the social media analysis. Further research should go into noise models. Deliverables: Short literature review of state-of-the-art methods used in price/numeric sequence prediction, detailed description of training chosen methods including a "how to" for incorporating sequence information taken from the web as well as the additional social media indicators into the model.
Overdue by 12 year(s)•Due by April 8, 2014•7/7 issues closed