The understanding of physics and physical processes and transformations makes it possible to identify precursors of anomalies, not just anomalies. Clearly, there are anomalies which are sudden and which do not produce precursors. In such cases, all one can do is to recognize the existense of the anomaly as quickly as possible and try to limit the damage. However, in anomalies having endogenous origins, i.e. rupture or malfunction due to wear and tear, software malfunctions – often triggered by sensor malfunctions – generally are preceded by events that may be used to forecast an anomaly.
The same may be said of medicine, a field of immense complexity. In a recently published paper, it is shown how QCT – our Quantitative Complexity Theory – can predict the response to a vasovagal syncope. In practice, the QCT can predict, even two minutes ahead of time, when a patient will faint, showing its superiority over conventional approaches. From the mentioned paper, the graph below shows complexity of patients with syncope and those without, indicating the predictive power of complexity and with a clear separation between the two categories. It is clear how rapidly increasing complexity precedes syncope already two minutes before it takes place.
Conventional approaches, such as Mean Arterial Pressure (MAP), as not as good in syncope prediction, see graph below.
The QCT, in practice, takes anomaly detection to a new level. Artificial Intelligence and Machine Learning have been around for a few decades. While ML works well in certain applications, we felt that it was time to develop a radically innovative and unconventional data analysis technology, redefining the concept of anomaly and anomaly detection.
In highly complex situations with hundreds of thousands of variables, Machine Learning is not feasible as it requires numerous examples of anomalies to learn from. Often you don’t have the time. Our QCT doesn’t require training or prior knowledge, only streaming raw data.
Reblogged this on muunyayo .