The detection of anomalies is a fairly common problem that covers many scenarios, from financial fraud to failures in a computer network. Some problems require complex models of machine learning, but most often some simpler and more cheap methods are enough. For example, there are data on sales for a period of time where it is necessary to celebrate days with abnormally high volumes or to allocate customers with an abnormally large number of credit card readings to check the risks. For such cases, a simple statistical method of release of emissions, called Z-scoring, is suitable. The assessment is equal to the difference in the current and average values, divided into standard deviation. Z-scoring involves a classic normal distribution of random values. The transformation of values in the nominal scale into a logarithmic scale will improve the ability of most ML models to distinguish between relationships and improve the ability of Z-indicators to note emissions.
#machinelearning #artificialintelligence #ai #datascience #programming #Deechnology #Deeuplearning #bigData #bigdata