What's Covered
This tutorial illustrates how to access match metrics. The following is covered:
- How to retrieve device Id.
- How to retrieve match method.
- How to retrieve difference.
- How to retrieve rank.
Code and Explanation
Match metrics example of using 51Degrees device detection.
The example shows how to:
const char* fileName = argv[1];
const char* properties = "IsMobile";
fiftyoneDegreesInitProviderWithPropertyString( fileName, &provider, properties, 4, 1000);
fiftyoneDegreesWorkset *ws = NULL; ws = fiftyoneDegreesProviderWorksetGet(&provider);
ws->method
ws->difference
fiftyoneDegreesGetSignatureRank();
fiftyoneDegreesWorksetRelease(ws);
fiftyoneDegreesProviderFree(&provider);
This example assumes you have compiled with 51Degrees.c and city.c. This will happen automatically if you are compiling as part of the Visual Studio solution. Additionally, when running the program, the location of a 51Degrees data file must be passed as a command line argument if you wish to use Premium or Enterprise data files.
Summary
Match metrics is an additional piece of information about each match result. It can help developers spot fake User-Agents and determine the general level of confidence the detector has in the current detection results.
Device metrics consists of four parts: device Id, detection method, difference and rank.
Device Id is composed of four components, each represented as a number and separated by the hyphen symbol. Numbers correspond to profile IDs the detector has selected for that particular component. The four components are: hardware, software, browser and crawler. For more information please see the 51Degrees Data Model .
The detection method provides information on what algorithm was used for this particular detection and difference indicates by how much the provided User-Agent is different to the best signatures found in the data file. The larger the number the less confident the detector is. Difference for the 'exact' method is always zero. Difference for the 'none' detection method is irrelevant. For more information see the How Device Detection Works page.
Rank value tells you how popular the identified device is. The lower the value the more popular the device. Popularity is derived by 51Degrees based on our observed usage level.
Other uses for match metrics include:
- Ranking devices by popularity. Best used in conjunction with other tutorials to rank the results based on the general device popularity.
- Storing device Id for analytics instead of individual profiles.
The Lite data file contains 446,634 device combinations, whilst Premium and Enterprise contain 940,018 and 1,407,838 device combinations respectively. A larger set of device combinations leads to the 'exact' detection method being used more frequently and the rank providing a more accurate value.