During the last months, I was focussed on doing Financial Forecasting using Neural Networks. It is really amazing how good these kind of problems could be solved with that.
If having enough actual data on time serieses and the business is relative stable, it is quite easy to get results with an accuracy up to 80% – 90%.
But sure, such results depends on the quality of the data. So I never had datasets which worked out of the box. There was always data cleaning processes needed to get a good quality like above.
So everone who thought, that there is no work to do anymore regarding forecasting using AI. Hi guys, unfortunately I have to disappoint you. It helps you to do less recurrent ordinary work and give you more space for creativity.
After reaching the first successes, the ambition was coming up to me. What to do to improve the accuracy? Believe me, there is a lot you can do. Imporoving the data quality, change the model configuration, adding new features, etc.
One of the faszinating points is to find and add new features which influences the predictions. It’s not only adding new values you have inside the data, but it’s more about filtering, calculation and manipulations or to find also external data. It’s like a game to find hidden secrets with limitless creativity. It’s an addiction in finding percent points improving the quality.
As Neural Networks are still stupid (it’s only an algorithm), I believe, this will be one intelligent works for people in the future. Robots will not kill our jobs, but the work will change.