Based on our patented technology, US Patent 9824060, the cbmLAD software is developed to help you to learn more about your process and the factors that affect its performance. This learning is based on knowledge extracted from the data you entered to cbmLAD. The data is usually gathered over a certain period of time which may be days or years. What is important is to have enough data that you think covers all possible states in which your process can go through. For example if you think that your process changes very rapidly, then a small number of successive observations may be enough to capture all the variability of your process, but if it changes very slowly over time, then you have to observe it for a long time, or have enough collected historical data that captures the variability in your process. In some applications the observations are collected from entities that are completely independent. In this case you will need data that represents all the states of these entities. In general, the more data you have, and the better this data represents real life with all its variation, the better your chance to capture the phenomena you are studying.

​

The knowledge, in its simplest form, is extracted and given to you by the cbmLAD software in the form of interpretable patterns that differentiate between classes or categories of observations in the targeted dataset. The classes and the phenomena are defined by the user. cbmLAD will discover the hidden knowledge from the data that you entered, which represents the factors that may be related to the phenomenon that you are studying. cbmLAD will confirm or refute the existence of such relations. As such, the results that you will get after running cbmLAD are:

 

• Interpretable patterns, which are rules found in your data, and which differentiate between the states of the system under study.

• Since the patterns are defined in terms of the factors’ values, the frequency of appearance of a certain factor is an indication of its importance in explaining the phenomenon under study. A factor that has never appeared in any pattern is obviously not important in explaining such phenomena.

• cbmLAD allows you to test the accuracy of the acquired learning in two ways:

  • By dividing randomly your input file into two separate files; the training file and the testing file. cbmLAD allows you to experiment with a different percentage of records in each file.

  • By letting you enter your training and testing files separately.

• cbmLAD allows you to classify new unclassified observations.

What is cbmLAD?