The Bitcoin prediction tool is a unique artificial neural network that predicts future Bitcoin price movements on Bitstamp every hour via a type of artificial intelligence; similar to a human’s central nervous system.
According to Wikipedia, There is no single clear definition of what an artificial neural network is:
However, a class of statistical models may commonly be called “Neural” if they possess the following characteristics:
– Consist of sets of adaptive weights, i.e. numerical parameters that are tuned by a learning algorithm, and
– Are capable of approximating non-linear functions of their inputs.Examinations of the inspired the concept of neural networks.
In an Artificial Neural Network, simple artificial nodes, known as “neurons”, “neurodes”, “processing elements” or “units”, are connected together to form a network which mimics a biological neural network.
In this particular case, the Bitcoin prediction tool is an artificially intelligent prediction model. Given the relatively the short time that BTCPredictions.com has been up, the Bitcoin prediction tool has an overall track record of being amazingly accurate.
Above each chart is the average error for the neural network prediction and how many minutes ago it was last updated. The 24-hour prediction chart averages around 1.2-1.3% error; although I’ve seen the error percentage as low as 1.117%. The 5-day chart averages around 2.6% error and the 20-day prediction chart has ~8.2% average error. Essentially, the further out the prediction, the less accurate it is. Once an hour (on the hour) the Bitcoin predictions are updated.
So how do we know that the predictions are indeed this accurate?
According to BTCPredictions.com:
Every time a prediction is made, the software making the prediction looks at the prices for all Bitcoin transactions in approximately the past 3 years. It practices making predictions on this data before it makes a real prediction. The software predicts the price at tens of thousands of points over the past three years and calculates the average error, so the average error for each chart reflects the average error of the neural network used to make those predictions.
Interestingly, the overall error percentage gradually decreases over time as more data becomes available to make its predictions. However, the Bitcoin Prediction tool Developer warns that major news events that cause a massive BTC price move also tend to (temporarily) cause less accurate predictions.
He indicated that he was currently working on a more advanced neural network to help reduce this problem. When contacted concerning how much more accurate it is, closer to now (i.e. the first three hours on the 24-hour chart, for example), he replied back affirming that shorter term predictions are more accurate with the specific predictions model that he’s using.
He went on to explain that the predictions for the next one hour or two hours are going to be a bit more accurate than the average error listed above the chart (on average); whereas the predictions for 23 or 24 hours into the future are going to be a bit less accurate (on average). He also revealed that he’s quietly working behind the scenes on 5-minute chart updates (vs. hourly, as it is now)!
Disclosure: I am not affiliated with BTCPredictions.
Featured image by Shutterstock.
Last modified: July 14, 2014 22:11 UTC