Over a million developers have joined DZone.

Astrological Predictions Checked by Machine Learning

DZone's Guide to

Astrological Predictions Checked by Machine Learning

You may be surprised to learn that in general, predictions by astrologers and by machine learning models are not very different.

· AI Zone ·
Free Resource

Did you know that 50- 80% of your enterprise business processes can be automated with AssistEdge?  Identify processes, deploy bots and scale effortlessly with AssistEdge.

Nobody can predict the price of Bitcoin. Recently, Standpoint Research founder Ronnie Moas revised his 2018 bitcoin price prediction from $11,000 to $14,000. A number of economists have made their predictions which didn’t come true. Mark T. Williams, best-known as Professor Bitcoin, predicted Bitcoin would fall down to US $10 by mid-2014. While the value dropped quite a bit in 2014, it never got even close to the US $10 point.

We conducted an experiment and used astrology for prediction. Even though there have already been astrological forecasts of bitcoin price, nobody has ever checked them with a machine learning algorithm.

We are:

  • Ekaterina Vasyanova, a graduate of the Scientific Astrology School by S. D. Bezborodov, is a certified analyst in human design (International Human Design School). Ekaterina has been consulting people since 2006 for astrology and since 2012 for human design. During this time, she analyzed more than 5,000 natal charts and 1,500 body graphs.

  • Timur Karimbaev is an IT expert with 15 years of experience in complex systems development. Timur is the co-founder of GetCourse, one of the largest Russian ed-tech services, and the co-founder of Human Discovery Platform, a system for complex methods of human personality analysis creation powered by blockchain.

Step 1: Building Astrological Forecast of Bitcoin Price for 2018

We took the date that the domain bitcoin.org was registered — August 18, 2008, 13:19:55 UTC — as a starting point. Astrology implies that every object (man, company, or animal) has a creation date and has its own set of attributes, and each day, you can superimpose the current position of celestial bodies on an object to get predictions for today. For example: “On November 16, 2017, Bitcoin has active gate 5,11, 32, 42, 43, 54, 57, which gives the activation of the channel 32-54 and solenoid in the channel 10–57.”

Image title

Here are the key points of the astrological forecast:

  • A strong decrease in March-August. Bitcoin will become the part of the global financial system. It’s expected that bitcoin and the world financial system will be interdependent and will impact each other. The state and the financial system will experience a threat from Bitcoin and will take various steps to restrict Bitcoin influence.

  • Increase in August-December. Legalization and taxation may be applied to bitcoin buying and trading.

Image titleA problem within the astrology is subjectivity of interpretation. All astrologic signs are explained by an expert from their own viewpoint. It brings a human factor into the prediction model, which has to be removed.

Step 2. Checking Astrological Forecast With Machine Learning

To see if the astrological forecast could be true, we checked astrological characteristics of Bitcoin price over the last three years and its actual price trends with machine learning.

To do so, we collected data about Bitcoin price for the last three years from an astrological perspective and linked them with its actual characteristics.

At the beginning, the source data looked like this. There are some notes corresponding to each date — the bitcoin price, the difference of the price with the previous day, gates, and some more data about the lines, channels, and magnets.

Image title

To prepare data for the model, we spread out all the channels and gates to the columns. If channel 14 was active on a particular day, the column will have 0, otherwise, it will have 1. We can compare the rates of the last and first dates of the month, and if it increased, the target value will be 1; if not, it will be 0. It would be logical to evaluate the rate within days, but here, we are checking the astrologer’s forecast, which is monthly, and we couldn’t build a sustainable model for each day.

Prepared data look like a big table with >150 signs and 2,600 lines (the number of days elapsed since the start of Bitcoin).

Image title

Then, we remove the month and date and check the accuracy of the model.

Image title

  • LogisticRegression accuracy: 0.788262370541

  • GaussianNB accuracy: 0.683544303797

  • SVC accuracy: 0.901035673188

  • KNeighborsClassifier accuracy: 0.940161104718

Let's check the classification quality — how well it predicted the rise and fall.

Image title

Using the K-nearest neighbor method, the model predicted the data on the test sample with an accuracy of 94%.

The model found a dependence between the astrological data and the direction of the price trends.

For each month, we evaluated the price at the beginning of the month, the price at the end of the month, the difference between them, and the probability of a price increase based on the model calculated.

Image title

Here's the prediction from the model:

Image title

It appeared that Bitcoin price astrological forecast was not that different from the ML model.


  • In general, predictions by astrologers and by machine learning models are not very different.

  • In February-March, where the astrologist predicts bitcoin price decline — the machine learning model gives a very fuzzy result. With this probability level, the model neither confirms nor refutes the astrological forecast. But unlike the neighboring months, there is no clear indication of growth according to the model.

  • In May-June, the negative forecast of the astrologer contradicts the ML model. It is the only place in the whole forecast where the model is strongly at variance with the forecast.

  • In July-August, the negative forecast of the astrologer neither confirms nor contradicts the model. Also, in September-October, the positive astrological projection does not contradict the model.

Consuming AI in byte sized applications is the best way to transform digitally. #BuiltOnAI, EdgeVerve’s business application, provides you with everything you need to plug & play AI into your enterprise.  Learn more.

bitcoin ,machine learning ,ai ,algorithm ,predictive analytics ,tutorial

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}