25 lines
1.4 KiB
TypeScript
25 lines
1.4 KiB
TypeScript
import { TimelineElementParameters } from '../../page/timeline/timeline-element/timeline-element-parameters';
|
|
import { Video } from '../../page/video/video';
|
|
import forexPoster from '../media/forex.jpg';
|
|
import forexMp4 from '../media/mp4/forex.mp4';
|
|
import forexWebM from '../media/webm/forex.webm';
|
|
import { videoPosterAltText } from '../shared';
|
|
|
|
export const forex: TimelineElementParameters = {
|
|
title: 'Predicting foreign exchange rates',
|
|
date: '2019 autumn',
|
|
figure: new Video({
|
|
poster: forexPoster,
|
|
mp4: forexMp4,
|
|
webm: forexWebM,
|
|
invertButton: true,
|
|
altText: videoPosterAltText,
|
|
}),
|
|
description:
|
|
'From the animation, we can see that my implementation does a somewhat acceptable job at predicting (blue graph) the EUR/USD rates (green graph).',
|
|
more: [
|
|
'In a nutshell, the algorithm (written in Python using NumPy, SciPy, and Flask) predicts in the frequency domain. The steps are the following: smoothing the input values, differentiating, applying a short-time Fourier-transformation with overlapped (and Hanning-windowed) windows, extrapolating and then applying the inverse of these transformations to the resulting values.',
|
|
'Of course, there is still plenty of room for improvement, but even with this simple algorithm a mostly profitable trading strategy is viable. In my free time I may put more work into it.',
|
|
],
|
|
links: [],
|
|
};
|