Midas Touch Twitter Data Project for Vela Partners Microinternship
The objective of this project was to iterate on an Investor Similarity Scoring algorithm developed previously and develop a tool which would allow for the computing of a similarity score between two investors who are active on twitter based on their likes and comments of a given users post.
Two functions were developed: n_nearest_neighbours and plot_n_nearest_neighbours. The aim is to identify and visualise the n incvestors who act in a way that is most similar to the investor of interest. The function n_nearest_neighbours calculates similarity through 4 different measures (one of which takes into account the number of likes that a given user has recieved). The plot of this function allows for observations of the similarity between investors and for the observation of whether or not investors are interacting with each other - and if they are, whether they are interacting with similar investors.
The Vela Twitter Midas file has a working algorithm. The Vela Twitter Midas Organisation file was an attempt to group investors into organizations and observe organization/investor similarity. This does not currently work - although perhaps trying to scratch code in multiple if statements was too messy and I lost track of variables on google collab...
The full project description is found here:
Full credit to J.P, whose algorithm and graph I iterated off of. Many thanks to Vela Partners for giving me the MIP opportunity.