CryptoPredicted | AI-Driven Crypto Price Forecasts Using Big Data And Machine Learning
It is every stocks trader’s dream to always make the perfect investments. Unfortunately, technology has not yet reached the stage of predicting future events in fluctuating markets. But we are able to increase our chances of making better trading decisions, this is exactly what CryptoPredicted does.
CryptoPredicted is an early stage startup with one core purpose: helping you to trade better and worry less. They provide a service based on pseudo real-time data pulled from hundreds of external sources. Apart from their ability to make pretty accurate short-term predictions, they allow you to look at data in completely new ways. Do not fear any complex formulas or indexes, because most of their charts are very intuitive, even a 12-year-old kid can read them.
On the general chart, you can toggle a wide array of graphs by clicking on the labels. Every x-axis tick consists of data aggregated for the selected interval (“1 hour” on our screenshots)
A few useful graphs which you can toggle include but are not limited to: sentiment analysis from social and news channels, trading volume of the past 24hrs, social mentions, and news mentions. These graphs allow traders to view many different indicators on a single graph and understand where the market is headed.
Another useful feature they provide is trendlines. Some graphs can be reshaped into partial trendlines with a variable size. On the image below we have selected 6-point trendlines and enabled two trendlines: social sentiments (in blue) and the average price (in black). As you can see both trendlines look quite alike, this means that Bitcoin’s price is partially influenced by the social mentions and/or vice versa.
The trendlines appear to be quite useful for detecting trends in the market, but also to see which types of data are related to one another. Below is an animated GIF illustrating how the graphs transform as we change their trend size. In this example, we have toggled on the average price (yellow), social mentions trendline (blue) and trading volume delta trendline (dark red).
Let’s move on to the predictions chart. This tool contains many more settings to play around with, but it’s recommended to use the default ones. Their predictions tool is based on advanced machine learning techniques. It’s still in an early stage but already produces some nice results.
On the predictions chart, there are three types of graphs: the black line is the actual average price. The brown lines are the future predictions, the red line is simply the average of the predictions. The green lines we ignore as they are solely for development. Another thing you notice is the signal indicators of when to buy and/or sell. These signals are based on the predicted average.
A very useful feature on the predictions chart is the ability to look at historical predictions. On the image below you can see that the price’s line overlaps the predictions region. This allows us to see how accurate a certain prediction was at some historical date and time. In the end, it all comes down to buying low and selling high, preferably with a very high margin. Just keep in mind that it is still an early stage tool, thus not every prediction will be accurate.
The reason why there are multiple brown prediction lines is that each of these was generated using a different combination of features. The predictions themselves are generated by means of training an artificial brain, also known as a neural network. This brain is trained by feeding it data as seen on the general chart. Basically, it learns how the price went up and down throughout time and taking into account many other features such as the sentiments, mentions, traded volume and more. We did notice that the short-term predictions (i.e. first 3 to 5 intervals) are more accurate than long-term ones (i.e. beyond 10th interval) — making this tool ideal for active day traders.
Crypto Analysis in 3D
If you are bored of looking at graphs all day, then try staring into their 3D plots and be mind blown. There are several 3D plots available right now, we will only look at two of them.
The first one is the “time-price-volume” plot, using 5-minute intervals with 6 days worth of data. What makes this plot interesting is that we can see how the price and volume change throughout time. It’s quite a different experience than looking at a two-dimensional plot of the same data. What’s even more interesting is that we can see a wave-like formation. It appears to be quite useful for making trading decisions because one can see when there is a turning point in the data, i.e. a trader can decide when it’s time to sell or buy. And since the data is quite cyclic we can visually detect patterns and compare them against historical cycles.
The philosophy behind this plot is that each cryptocurrency can be categorized by a certain stage. The stage of a crypto is determined by its popularity and worth. What Bitcoin was worth in 2013 is not the same as what it is worth right now. Their idea is to categorize the coins and track their evolution throughout time. This allows us to see which coins are more mature and stable than others. But it also allows us to see when a certain coin has reached a new phase relative to other coin(s).
Overall CryptoPredicted is making fast progress, considering the fact that they started their development in late December 2017. And in less than a month already released a public beta which you can freely access here: http://cryptopredicted.com .
On first sight, the predictions seem to be pretty accurate for most intervals. So that appears to be their most valuable asset. Since their service is in public beta phase it is a golden opportunity to make the most of it.