Clarifai is an artificial intelligence API that recognizes image and video inputs. By generating highly accurate tags that links to the image, we are able to qualitatively describe and categorize the visualized information. We therefore notice that there exists great potential in using such information for correlation and trend analysis to discover results that are not as observable in daily life.

What it does is a web-app that allows users to input a keyword and find category tags that have the strongest correlations with it. Specifically, our algorithms dig into online picture databases for images based on the keyword, look for the tags information using Clarifai, and return the correlations in the form of heatmaps. A rank of the different correlations based on their strength with the keyword input and its synonyms will be available.

One example target user might be a designer who is trying to determine which color is the best to can use for an abstract art piece that represents success.

How we built it

We used as our platform, html & CSS for the website programming and js.node to implement the APIs. The APIs we used are Words API, Clarifai and Microsoft Cognitive Service API.

Challenges we went into

We started with the idea of generating heatmaps for equity and portfolio analysis using the Bloomberg market API. Our original design wished to provide users a plethora of functions, including valuation methods comparison and portfolio analysis based on factors such as region, market cap, and risk adjusted return. However, we later found out that the API is no longer available. As we tried to find alternative APIs that could provide us live and past market data, we realized the impracticality of using data sets that are too large and constantly changing (such as stock market data). Therefore, we moved on to the idea of implanting searching function in our design.

Accomplishments that we are proud of

Although the four of us have very different hacking backgrounds, we have managed to divide our work in a highly efficient way such that our strong suits in front-end, back-end, design, and presentation could be fully brought into use. We also overcome the technical shortage in data sets and APIs, and even improve our design of the heatmap generator based on earlier difficulties: by applying the Bing AI API in our code to equip it with fuzzy search function, we enable our website to provide stronger and faster correlation/trend detecting power

What We Learned

We learned how to quickly manage and use APIs, and how to make our algorithm more efficient. We also learned to exhaust every possible mean in order to solve one problem: there was a moment when we even tried to generate code using code, by printing some code to the console and copy it over.

What’s next for

Although the Bloomberg API was down, we would still want to add our feature of valuation methods comparison and portfolio analysis. By creating a friendly user-interface and accurately updated stock market data, we hope to gather a significant user population of equity researchers, traders, and other financial professionals. In order to secure our users, we will further advance our algorithm to allow for more functions for users to choose from. As more and more functions come into place, we plan to created a membership system with a monthly subscribing fee to our website in order to unlock the full functionalities.

For the open-ended correlation function, our focus on the user experience and the range of the website’s service still apply. In order to secure our users, we would strengthen the functionality of our heatmap generator by creating more specific algorithms with industry focus to allow users to execute more powerful search of correlations and trends. In addition, we plan to allow users to upload their own photo library and receive personalized generated correlation/trend report based on the library.

Built With

Share this project: