Public opinion, publishing house, institute of fund (Deductor)

Task

The fund conducts regular polls on a broad spectrum of questions in the majority of territorial subjects of the Russian Federation.

The greatest interest is attracted by the moments of sharp change of public moods which are difficult for detecting in a huge data stream of sociological polls. With respect thereto the task to develop and test a technique of automatic identification of anomalies in data of opinion polls, with assessment of degree of deviations was set.

This information is of essential interest since allows to reveal quickly the moment of change of public moods and to consider them at decision making.

Solution

BaseGroup Labs analyzed different combinations of questions of questionnaires and the studied objects: batches, persons, territorial subjects of the federation and other.

Use of algorithms of robastny filtering allowed to select automatically timepoints when significant change of public opinion was registered. Thus, the rating of subjects to which it is necessary to pay attention first of all forms. One of the important criteria influencing anomaly assessment is difference degree from a situation in other regions and over the country in general.

Interesting feature of the offered technique is capability to self-training, i.e. to automatic fine tuning of the selected abnormal thresholds at emergence of a new portion of data.

Results

The Public opinion fund is going to use the received results in the activity for increase in level of information filling of the reports provided to clients.

Source

http://www.basegroup.ru/clients/fom/