Empirasign Blog Entry: Collaborative Search
Collaborative SearchOct. 9, 2018
Utilizing an item-based recommendation system similar to those employed by Amazon and Netflix, Empirasign now suggests other bonds that you may be interested in based on BWIC contents, proprietary aggregated search history, and site traffic from our worldwide user base. These search recommendations are self-organizing and don't depend on any outside reference data, market levels, computed or theoretical cash-flows. The only input is Empirasign users' collective actions. Such recommendations, when available, display on all single bond search results pages. Example screenshot below.
How does it work?
The algorithm is relatively simple (albeit computationally intensive) and largely based on a retrospective of Amazon's recommendation system techniques published in the IEEE Internet Computing journal. That being said, the secret sauce is really the data. Very few players in the market have the collective and real-time aggregated bond search history to calculate such metrics and recommendations. Even so, these algorithms are such data hogs and the number of Structured Product bonds is so high (about 3,000,000) relative to the number fixed income market participants, a collaborative algorithm can only generate recommendations on about one-half of all searches.