Newbie in data science designing a model which predicts temperature/stock value

External question — Data Science Stack Exchange Author: Root Groves Original post: https://datascience.stackexchange.com/questions/138010 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. Suppose we had a array of points(let that be stock value,temperature etc...).If we did the DFT of such a array we could find the frequency spectrum and could predict the next value of the sequence. However what if we had another sequence which told us what the actual values were.Now we could check if the prediction matches with the current value and if it is not we have to correct the prediction. So I was wondering if we had a coefficient $\eta$ which described this and $\eta * c_{dft}$ becomes the new value of the predicted sequence.At the beginning we start with $\eta=1$ so if the DFT tells us that the new predicted value was lets say 5 then we get 5 as the first predicted value and if $\eta$ is increased or drops depending on the actual values then the predicted value would be $\eta$ times the coefficient calculated by the dft.What is a starting point for $\eta$ and how much should $\eta$ be optimally influenced by mismatches between prediction and reality.Is there a standard for this?