I try to be updated about A.I. even if the topic has deflagrated in the latest years and is now a rapidly growing bunch of niches. So, recently I used a public domain website to generate the content for a post starting from the title “Using Artificial Intelligence in Your Trading Strategy to Maximize Returns”. 🙂

After the intro and under the subtitle “Know your data”, this paragraph:

A key concept in AI is that the more data you have, the smarter your algorithm becomes. This means that you need to input as much data as possible before starting trading. You can use any information that can help with your predictions, like financial news or social media updates.

Maybe the a.i. is deliberately trying to kid me, which means that Skynet is already at work, or we have to consider the fact from another point of view.

First: the statement “the more data you have, the smarter your algorithm becomes” is partly true, but basically wrong. You MUST have the correct data AND a well-designed training procedure and THEN the abundance of data can produce a more brilliant output (aka a lower quantity of errors).

While input data can be noisy and even somehow slightly malformed, the training hall must be perfect, because here happens the magic and the meaning is associated with the input.

The input is there, no way, but the meaning is somewhere else. Is it in the chart? or in the picture? or in you?

Patterns form in your mind before any neural network and this side of the procedure is what makes the difference. Evaluating any A.I. task, knowledge and experience are fundamental.

Coming to the analysis of time series, I see a lot of production and methods to forecast a time series from itself. Those are boys, knowledge and experience say that you never have on the right what you have on the left. If as an output you want any kind of classification or pattern recognition, that classification or pattern must not appear on the input side.