New Forecast. Updated How it works Client Application / install it

Prediction is very difficult, especially about the future.
Niels Bohr
Premium Markets stocks and shares price trend forecast based on neural network, financial technical analysis and sector rotation

Updated from the first 2012 post.

Premium Markets, Free stock market Forecast, Trend Analysis, Investment Advice, Buy Sell signals, based on Sector Rotation and Technical Analysis

Google adSense Privacy & Terms

The stock market trend forecast engine

Premium Markets is a free stocks and shares price trend prediction engine base on neural networks. The stock market forecast is solely based on technical analysis of individual stocks or markets. Demoed on this site.

This free trend forecast software ought to help you, as a stock market beginner or as an advanced investor, in stock market screening, performance projections, market analysis and ultimately in finding the best future stocks and shares for your swing or long term investments. Premium Markets offers stocks buy sell signals monitoring via buy sell notifications sent via email.

The stock market forecast engine is available as a free proof of concept and demo online. It also is available as a standalone application. The latter includes additional features like price historical data charting, trends monitoring, financial technical analysis, indicators editing and stock portfolios management.

Please don't hesitate to enquire if you are interested or want to know more.

May Premium Markets be a free addition to your stock market watch and stock market trend analysis.

How does it work?

If the next price of a share may be impossible to predict in an accurate manner, it may, however, be possible to forecast the trend a share price will take in its near future. Knowing the trend, one can aim at buying near the lowest price and selling near the highest price over a defined period of time. Premium Markets offers a forecast engine based on historical data analysis. The engine aims at finding what the next trend of a chosen market, a specific share or a set of shares is/will be. In other words if the subjacent is bullish or bearish.

Using Financial Technical Analysis

The main issue faced using technical analysis is that it seems difficult to master without many experiments, back testing, practice, knowledge and as many failures. Difficult, indeed, to find out in an accurate manner what a signal or a set of signals really means regarding the trend a share price will follow. The Premium Markets forecast engine is here to try and answer this issue and give everybody the ability to use technical analysis without necessarily knowing that much about it.

Principle of the calculation:

When back tested on historical data, the calculation results give us a curve similar to the desired output, this will be an SMA in the following illustration. In reality, the SMA, calculated with live data, is lagging behind the quotations. This makes it unusable in a volatile market.

Our prediction is not.

This is fairly amazing considering that we don't directly use the historical data as input, i.e. we don't need the actual SMA nor the stock price as input, but just signals and calculations data derived from the daily market data. Hence somehow we predict the future SMA of the stock price without direct knowledge of either.

Comparing the prediction against Moving Averages (MAs) like SMA and WMA

To get a better understanding, in the following charts we compare the forecast output out of the infamous BTC-USD with the SMA and WMA 50 applied on the BTC-USD prices.

Price trend forecast chart of the BTC-USD based on technical analysis with lagging SMA and Weighted SMA 50.

The dark blue line is the stock price.
In yellow and pink are the SMA and WMA 50.
As you can see, and this is well known, both MAs are lagging behind the real trend (of about 25 days).
A longer SMA period gives a better match with longer trends and a shorter one with shorter trends. But both are lagging.
As you will see in the next chart, the neural network output calculated by Premium Markets aims at answering these issues.

Price trend forecast chart of the BTC-USD based on technical analysis with no lag SMA and Weighted SMA 100.

In this chart, we have 'fixed' the MAs lags by artificially shifting them 25 days in the past, letting the resulting MAs to display a better representation of the trend indeed.
The issue is now that, by shifting left, we are missing the most recent and important 25 days... This is what we call the lag.
In comparison, the output calculation from Premium Markets solves this issue. The light blue line is the trend line prediction issued from the forecast.

  1. The prediction redraws the shifted MAs with even seemingly a head start.
  2. The trend is similar to the selected MAs when their lag is fixed.
  3. The generation of false signals is mitigated with a smoother curve.
  4. And the trend is drawn until the last day of quotation available.

Bullish or Bearish: buy and sell signals

As seen above, the output is in fact drawing a smooth line of the stock price trend but without the lag that you would actually get with other smoothing techniques. By redrawing a smooth trend line of the quotation line, the output from the neural network will tell us if the share is following a bullish or a bearish trend.

Price trend forecast chart of the BTC-USD based on technical analysis with buy and sell signals.

In this chart, the green and red bars draw an interpretation of the trend line indicating respectively bullish and bearish trends.

In the top bars, we have the predictions and in the bottom bars, the targets.
Here the interpretation is simply based on the trend line going up or down. Its reversals used as buy and sell signals.
In this particular example, the forecast enabled us a good 10% profit as the stock price goes down -60%.
It may be worth mentioning that a finer interpretation can be made for each stock. For instance considering the slop and velocity, adding some delay when reversal occurs or using stop losses.

About the Neural networks at play

The concept was experimented with different neural network models and prototypes, mainly wavenet like models with 1D input and convnet like models with 2D input. In the case of convnet, the input is seen as a 2D coloured image. In the case of the 1D approach, as a time window.
As neural network engine, we used in the beginning the, then excellent, Neuroph and Encog, and now TensorFlow.

On this site you will can also find a live demo of Premium Markets trend prediction concept. Premium Markets also offers an email notifications service for trend and market watch.

If you are interested or want to know more: contact.
Dear Reader, this web site is a non-profit initiative. It is free. It aims at helping You in making profitable stocks and shares transactions. Give as ... every penny/cent counts and You can help.

Hosting and developing a cloud based web site has costs. If you found this site of some interest, take a minute to help keeping it online and improve.

Using Other historical data, Indices/Indexes Performances and Sector rotation.

The following few chapters are additional exploratory works dating from 2012 and have not been updated since. You may want to skip directly to the next section.

This feature is based on the sector rotation theory which in short aims at predicting stock markets trends based on price movements of sector indices. Following the same principle as for predictions based on technical analysis Premium Markets aims at forecasting the next trend of a share using historical data of these sectors indices and other major indicators. Again the historical data are injected into the neural network during a training phase. The latest data are then used to forecast the near future of a random stock of your choice. Here again, the aim is not to find the price value but more to predict the price trend.

On the following charts, you can see two sample results of a back testing against the French CAC 40 index.
Price trend forecast chart of the CAC 40 ending on May 2008, based on sector indices rotation Price trend prediction chart of the CAC 40 ending on February 2008, based on sector indices rotation

The dark green line is part of the CAC quotations used for training and calculation.
It respectively stops on May 2008 and February 2009 with a horizontal line.
This horizontal line represents the period of prediction (200 calendar days on these samples).
The light green line is the full quotation of the CAC. It is used here as a mean of comparison as it won't be used in the calculation, these being future data in the context of the back testing.
The red line is a smoothed representation of the resulting forecast output.

You can hence see the accuracy of the resulting forecast, over the 9 months of prediction, by comparing the red line to the light green one.
For a preview, developed as a proof of concept please enquire Contact
Also check Premium Markets Trend prediction based on technical analysis.

Global forecast based on technical analysis

Premium Markets also offers a global trend watch and forecast ran daily over the main stocks and shares markets. This stock market prediction summary gives a global view of the markets and industries current and future trends.
For a preview, developed as a proof of concept please enquire Contact
Also check Premium Markets Trend prediction based on technical analysis.

Portfolio management, the Standalone App

Premium Markets also offers a multi platform downloadable Standalone Application User Interface for portfolio management including historical data charting and technical analysis indicators monitoring. In its advanced version, the User Interface can also be used as an entry point to the forecasting engine.
For more details and downloads, see Premium Markets at and Premium Markets Standalone App.

Technical Analysis Indicators and Oscillators

In its current version, Premium Markets implements and uses more than a hundred indicators and oscillators, including :

For more see also

For more information about the calculation please refer to Premium Markets.
Or ask me at contact.