Weak signals are a crucial in the effective analysis of complex data, but due to the very nature of weak signals, they are not an easy thing to interpret.
What are Weak Signals?
Weak signals are changes that have ambiguous interpretations or origins.
These changes could be anything from a blip in a time series to an increase in the number of purchase orders related to a project. The inherent ambiguity combined with this wide range of possible sources poses an interesting challenge in analysis, especially when creating automated algorithms. However, weak signal analysis is something that we encounter regularly in everyday life, and we successfully interpret and act on them without even realising it.
As an example, take the case of a doctor who is visited by a patient exhibiting three symptoms:
- They have episodes of dizziness.
- They are sometimes short of breath.
- They struggle with exercise.
In isolation any one of these symptoms could be attributed to a great number of conditions. Dizziness could be caused by migraines, reaction to medication, or an inner ear infection. Shortness of breath could be an indication of an infection, bronchitis, or asthma. Struggling with exercise can be related to poor diet and fitness, respiratory conditions, or musculoskeletal disorders. Each of these symptoms is a weak signal.
Although each symptom could point to many diagnoses, the doctor is unlikely to start with any of the disorders listed above. This is because, when considered together, these symptoms are a common “red flag” for heart disease. This does not mean that heart disease is guaranteed to be the cause, but it is likely to be the first thing the doctor would want to rule out, meaning that the patient will be sent for an ECG to check heart function before they are investigated for bronchitis. The doctor has performed weak signal analysis to evaluate which disorder is the most probable cause of the symptoms.
The three symptoms are weak signals which, when taken together, indicate a significant probability of heart disease.
The same is true for any weak signal interpretation. The goal is to determine the most probable causes of the change, taking into account all of the available information. The action taken is determined by these possible causes, and their associated risk.
What are the Challenges?
Weak signals are not always easy to spot, the problem of separating weak signals from noise is not always easy to solve. Once identified, determining their probable causes is the next challenge, the complexity of which is dictated by the quality and amount of data available to build a picture of the processes in play.
In the example, the doctor knows to screen for heart disease because during their training experts have told them the links between these symptoms and the disorder. The experts know that the link exists because over the years hundreds of thousands of patients have developed heart disease, and their symptoms have been analysed and correlated to identify the relevant indicators.
The same can be done in industry. Over the years many similar plants have distilled crude oil, produced ethylene, or manufactured chemicals. Similarly, many projects of varying sizes have been completed, some on-time and within budget, some late and over-budget. In all of these cases data exists that can be analysed and correlated, and experts have developed an understanding of the underlying processes. This information can be aggregated to identify and analyse weak signals, providing intelligent systems that allow Sabisu to remain ahead of the process and project.
We’re always interested in hearing from you with any comments or suggestions, feel free to get in touch.