20Day-Ahead Prediction for FB Using Artificial Neural Networks

We have generated a Deep Learning Algorithm for predicting the future prices of stocks/commodities/foreign currencies/crypto currencies. Our algorithm uses sophisticated deep learning/artificial neural network approach for time series analysis.

Time series data is processed using self-adapting and self-optimizing algorithms, and is fed into the Neural Network. The network is then trained, and the multi-step ahead predictions are generated.

The algorithm is a complex of a number of different mathematical approaches, which processes the time-series data for any stock/commodity/foreign currency/crypto currency.

You can see the 20 Day-Ahead Prediction for FB             on the chart. And we will keep posting the Week-Ahead/10Day/20Day-Ahead predictions for different tickers. Please note that the outliers have been filtered out by the algorithm, therefore the results are indicating the moving direction of the price, which passes through the significant points given on the predicted results.

Your comments and questions are valuable to us, please share them below. And please don't forget to follow our channel on TradingView and on Twitter @AhmetSBilgin
Hi, thank you for sharing. As much as it sounds weird for many, I believe this is possible to some extend. I had this Idea few years back. Few of my friends and I started working on it to mathematically look at past behavior of the stock and factoring seasonality and even the tensions in the world to do the same prediction. However our focus was on predicting only 8 to 12 bars from 30 minutes charts to daily. It got too complex and everybody eventually left the hubby project. This was about 5 or 6 years ago, it is great to see someone is working on such idea.

Questions: have you back test it? Do you factor in the news and economic numbers? Do you consider the related underling assets, sector, and commodities. For instance, if you want to predict the GDX, you need to factor in the gold price, interest rates, and overall strength of the markets around the world. Or instead of all of these, the algo just looks at the technical indicators and assumes all other elements are priced into the chart? For a test, can you run your algo on the GDX and the Gold separately to see how the results compare?

Thank you.
AhmetSBilgin marjmandi
@marjmandi, Hi,

The algorithm has been tested for couple of months already, for different time intervals of different stocks. A part of the time series data has not been fed into the network, and kept as a test set.

We have compared the error values between the predicted results and the test set. The algorihm was improved accordingly.

Any relative time-series data can be fed into the network, such as other correlated stock/commodity prices and/or sentiment data, and the algorithm adapts itself to the time series relation between these factors to make the prediction.

Please note that, the algorithm is still in improvement phase, and additions/changes are still being implemented into the code.

marjmandi AhmetSBilgin
@AhmetSBilgin, Great, I'll keep an eye on your posts. Good luck with the project.
ZH 简体中文
EN English
EN English (UK)
EN English (IN)
DE Deutsch
FR Français
ES Español
IT Italiano
PL Polski
TR Türkçe
RU Русский
PT Português
ID Bahasa Indonesia
MS Bahasa Melayu
TH ภาษาไทย
VI Tiếng Việt
JA 日本語
KO 한국어
ZH 繁體中文
首页 股票筛选器 外汇信号搜索器 加密货币信号搜索器 财经日历 如何运作 图表功能 网站规则 版主 网站 & 经纪商解决方案 插件 图表库 功能请求 博客 & 新闻 常见问题 帮助 & 维基 Twitter
个人资料 个人资料设置 账号和账单 我的支持通知 联系支持 已发表观点 粉丝 正在关注 私人消息 聊天 退出