The nature of the programs are statistical. They are different from typical AI programs. This is not a new type of AI.
There are many algorithms to analyze data. My programs are designed to analyze time series data that are very random.
Although the programs have been developed to predict lotto numbers (too random), they will probably be more useful to data analysts. They may be useful to predict share prices, to decode signals from outer space, to test random number generators, to test any data for randomness, etc.
The programs are general tools, not specific solutions. It is up to you to figure out how to interpret their reports for your application. (It is not simple. It is fun if you like to play around with numbers)
A list of statistics is not very meaningful. To get a better perspective of what the reports mean, it helps too compare it to a benchmark. A report is also calculated for data from a true random number generator, a pseudo random number generator .
The program looks for
all occurances of sub sequences of numbers, n1, n2, ... nm
in a main sequence, n0 ... mz
where n1< n2< ... nm
and then calculates the probability of each sequence.
The goal is not to find a specific sub sequence, but to find downward trends.
The program can be used to predict the next number:
E.g. If you have observed that the last 3 numbers were: n1< n2< n3
then the program will tell you the probability that the next number in the sequence will be n0.
You can also specify a minimum value for n0.
E.g. If you want to predict when a share price will drop low enough to make it worth buying.
The program can be used to find underlying cycles:
If data is smoothed with some algorithm and plotted, then it is easy to see peaks and valleys.But to calculate the periods more precisely is difficult especially if the periods are not exact and if you are using big datasets.
The program will calculate a numerical value for the periods.
Caveat 1:
Sequences, especially the shorter ones, will overlap:
E.g. The sequence n1< n2< n3< n4< n5 will have:
patterns of 3 numbers: x3
patterns of 4 numbers: x2
patterns of 5 numbers: x1
Caveat 2:
If the data has a long term upwards or downwards trend, the results may be meaningless or unreliable.
For a free report, email a list of integers (no more than 5000) each on a separate line
to: HeinRich@BelleModels.biz with subject: Finding patterns in random data - Drops - overlapping
Although it is not a requirement, I shall appreciate it if you could tell me what you want to use the numbers for.
Spikes are an easily identifiable feature. A spike is a data entry that is significantly bigger than the average.
The program calculates
The program can be used to predict the next number:
E.g. If you have observed a spike
then the program will tell you the probability that the next number in a specific range.
You can also specify a minimum value for n0.
E.g. If you want to predict when a share price will drop low enough to make it worth buying.
The program can be used to find underlying periods
Caveat 2:
If the data has a long term upwards or downwards trend, the results may be meaningless or unreliable.
For a free report, email a list of integers (no more than 5000) each on a separate line
to: HeinRich@BelleModels.biz with subject: Finding patterns in random data - Spikes
Although it is not a requirement, I shall appreciate it if you could tell me what you want to use the numbers for.
