Interesting AI project to predict quarterly earnings

Artificial intelligence is making great strides, and there are many passionate people working on it.

An interesting project is the investment algorithm of puzzletolife.com, which uses neural networks to predict the movement of some stocks on the days this stocks announce Quarterly Earnings.

As you may know, American companies explain their quarterly results 4 times a year. The consequence of these announcements means that sometimes the company’s stock rises or falls sharply if it exceeds analyst expectations or falls if, on the contrary, it disappoints them.

Puzzletolife makes a Quantitative Analysis of the stocks, using several neural networks makes a prediction the day before the earnings are announced, trying to predict whether the results will be a positive surprise or not.

The team has told us that they use various types of neural networks (assembled in stacking), some aimed at analyzing the intrinsic value of the company and others to predict specific aspects such as volatility.

The team has been testing different NN architectures for years to find the best option. Among them they have tried types of reinforcement learning and even some oriented for images like CNN.

How to believe the results?

To build trust with the system, puzzletolife uploads the predictions of each day hashed with sha256 to its ethereum account, where they are recorded with a timestamp, to prove that they predict the day before to the earnings day.

We are going to explain an example to prove that the predictions they have are true.

We will pay attention to the bearish prediction OXFD for May 5 2020, if wehash [‘OXFD’] with sha 256 it comes out:

b68e026668f15c292b8dbb42d3bb5f68d74fd1e0249ad9105437a915cc707936

Which was recorded in this ethereum blockchain transaction on May 4 https://blockchair.com/ethereum/transaction/0xc7f329f23a15fb0e88185c0fe89d2a62eec53822e827d1d270c3982b8f59ef23

 

 

You can see it in the “input data” field of that web (it can be open up), and you will read it in the “down” part of the json.

The good thing about the blockchain is that you can check the transactions content for yourself, even downloading the blockchain, or creating your own transaction inspector. If you don’t want to do all that, you can also use different websites.

Another is this:

https://etherscan.io/tx/0xc7f329f23a15fb0e88185c0fe89d2a62eec53822e827d1d270c3982b8f59ef23 to see the input data on it (click view more, then view as UTF8)

On puzzletolife website you can check the results provided by the system, the team says that covid crisis volatility has altered stability and profitability, but they have several categories of predictions and provide information that can help a trader’s analysis, the website is not a financial advice.

To finish just add that the cost of watching the predictions is a subscription of 39.5 $ / Month. To be a subscriber also allows you check performance month by month.

 

This is a Sponsored Feature.