Data analysis software for fast, graphical data processing - Star-Oddi

PatternFinder

Graphic Data Analysis Software

Overview

PatternFinder is a graphic data analysis software designed to save the users resources by making data analysis faster, more efficient and easier.

Description

As the name indicates, PatternFinder can be used to find patterns (defined or undefined) from data, but there is more to PatternFinder than just finding patterns.

The main analysis features available in PatternFinder are:

  •  Periodic statistic analysis of time series data
  •  Signal filtering and operations
  •  Overlaying and analyzing multiple signals
  •  Event analysis
  •  Pattern analysis (pattern finding)

In PatternFinder you can further analyse all data from Star-Oddi loggers, but also data from other data logging systems. PatternFinder works seamlessly together with Star-Oddi supporting software’s SeaStar, Mercury and FoodStar, where retrieved data can be transferred over to PatternFinder with one button click. Other text format data (where you have “one measurement per line time series“) can be imported to PatternFinder for analysis by defining the data format. In PatternFinder you can analyse data from several sources at the same time.

In PatternFinder you can create Single File Projects (SFP) or Multiple File Projects (MPF) by selecting one or many data files to work with. You can choose data from different sources (Star-Oddi data as well as outside data) to be added in a Multiple File Project. PatternFinder is an ideal tool in processing large amounts of data that otherwise would take a long time, e.g. data from several tagged animals over a time period of several years.

All operations in PatternFinder are viewed graphically, but all results can also be viewed as XML files or text files. All operations and results are registered, can be renamed, viewed as reports and printed out. The history of all operations and their results for each signal is registered and can be viewed simply by clicking the history button found in all graphs.

Almost all operations in PatternFinder result in time series data, including statistical analysis results, and can be exported in TXT, XML, HTML table or Excel file formats to other applications. All Event and Pattern graphs have a single click “export” button that exports the results to text file.

Online Support

Click on the links to view the PatternFinder user manual, tutorials and case studies.

Features

Periodic statistic analysis of time series data

The user defines a period (time unit) the data is to be analysed in (i.e. months, weeks, hours, minutes, seconds). PatternFinder provides the user with a eleven signal graph including one signal for each statistic parameter: maximum, minimum, mean, standard deviation, median, range, points, energy, kurtosis and skewnes. Each statistic parameter signal can be extracted and saved as a separate signal and used for further analysis.

Signal filtering and signal operations

In PatternFinder signal stands for single data graph, whether a statistic parameter signal as described above or a measurement signal. Signals can be filtered before further analysis. Besides being able to “cut“ the signals numerous other filtering options are available, such as single spike filtering, smoothing, moving average, round off and scaling. These filtering options come in handy when filtering out “noise” or preparing signals for event analysis or pattern recognition. Signal operations include subtracting one signal from another and comparing two signals.

Overlaying and analyzing multiple signals

Multiple data signals (from one or many sources) can be viewed in the same graph. The data can be viewed as it is, shifted to start at the same time or as annual data. Interpolation of the signals is also possible (giving same value for all signals).

Statistical analysis on each signal type (same unit value) is possible, resulting in a eleven signal graph. In this case all the statistical parameters refer to comparison of all the signals in a singular point. Each statistic parameter signal can be extracted and saved as a separate signal and analyzed further.

Event analysis

Two types of event analysis are defined in PatternFinder, behavioural event analysis and ambient event analysis. In behavioural event analysis changes in the signal are defined. This feature is designed for analysing signals from tagged animals that indicate a behavioural signature. Signals presenting parameters such as pressure, magnetic field strength and tilt can be further analysed with behavioural event analysis.

In ambient level analysis changes in level or level status are defined. The ambient event analysis is designed for environmental parameters such as wind speed or light. Both analysis methods produce an Event graph (ESX) that can be combined in a multi event graph.

Periodic statistical analysis on events is possible, resulting in a eight signal statistics graph. Each statistic parameter signal created can be extracted and saved as a separate signal and analysed further.

Pattern analysis

PatternFinder can be used to find patterns in data either by detecting occurrences of user defined patterns or scanning the data for possible patterns. The user defines a pattern that can be saved as a pattern template. Using autocorrelation and enveloping, the PatternFinder finds all occurrences of the defined pattern in the data (signal). The user can define in the software how similar (%) to the pattern template the accepted pattern has to be. Results are presented in a pattern graph.

When PatternFinder is scanning data for possible patterns without user defined model pattern, the pattern search is based on Events. Firstly the data is prepared for event analysis (e.g. smoothed) and then pattern recognition is run on the event graph finding all possible patterns in the signal. As you can imagine, all patterns is a lot of patterns. It is necessary for the user to set up some criteria for the pattern filter. Results are presented in a user selection pattern graph.

Periodical statistic analysis on the patterns is possible, resulting in a eight signal graph. Each statistic parameter signal can be extracted and saved as a separate signal and analysed further.