monalysis trend detection in time series
Published: Sep 16, 2024 by monalysis GmbH
We have extended maviz with a new function for trend detection in time series. This function makes it possible to identify patterns and trends in time-ordered data, be it in damage, sales or production data.
With trend detection, you can detect linear trends, seasonal fluctuations and other patterns in your data. The feature is designed to provide you with precise insights to make informed decisions and better predict developments.
Trend detection has been developed to be modular and easily customisable to your needs.
In this blog post, we explain how it works and how you can use the feature in your processes.
How it works
When a time series is selected, predefined analyses are activated immediately, which are defined individually for your use case. You can then customise the predefined values and apply them to the selected time series. These adjustments are controlled via the side menu. It is also possible to display all detections simultaneously.
You can currently choose from the following recognisers:
- Moving average
- Rises & falls
- Minima and maxima
- Consecutive extreme values
- Value ranges
Moving average
The moving average is a central tool in our trend detection. By adjusting the value range in the side menu, you can influence how many data points are included in the calculation of the average at the current point.
Moving average with 100 data points
Moving average with 10 data points
The more data points you enter, the closer the moving average will be to the actual time series. It is almost impossible to find an optimum setting. For this reason, the number of data points is optimised individually for your data.
Rises & falls
The moving average is used to recognise rises and falls in the data.
You can influence and refine the recognised ascents and descents both by adjusting the moving average and by using the sliders in the side menu (see image below).
The recognised rises and falls are marked dynamically in the time series with the help of drawn areas.
Minima and maxima
Another important tool in trend detection is the recognition of minima and maxima. The various extreme values are plotted dynamically on the time series. There are further adjustments in the side menu that relate to the minima and maxima. These are:
Maximum height of minima & Minimum height of maxima
Minima or maxima are only plotted up to the height entered. This function is particularly suitable if only the lowest/highest extremes are to be marked.
Distance between the extremes
Extreme values must not be too close to each other. If two extreme values are within the specified distance of each other, only the higher/lower extreme value is marked.
Buffer to the moving average
The extreme values must not be too close to the moving average.
These adjustments can be used to reduce the plotted extreme values so that only the points of actual interest are displayed.
Consecutive extreme values
The function for recognising consecutive extreme values is specially tailored for vehicle data. Pairs of consecutive minima and maxima are to be labelled. This can be shown particularly well in the trend detection in the display if both minima and maxima as well as the consecutive extreme values are activated.
Value range
A general range of values can be displayed in the time series, which can be used to recognise irregularities. When initially activated, the value range is first calculated using the average. If a more precise definition is required, the limits can also be entered manually.
When using the report function to export the time series images, the trends activated at that time are also displayed, which can be very helpful when communicating interesting points for the analyses.