The product nkd is given in millions, and z*, have been rounded to integers. Computing times t, to are in seconds for the small instances, and in hours for the large ones.

Summary of all notations in this paper.

Results of O-K-means for still larger instances.

Sample Twitter popular keywords and tweets associated with normal social dynamics.

(PDF)

Information to seize the relationship between computational effort and solution quality.

Correlation coefficients of synthetic and real instances.

The models used include VARMA and SVR, whereas the features used include the raw social dynamics (Baseline), level-1 activation vectors (H1), level-2 activation vectors (H2),...

Simulation input parameters for the manipulation study.

Some results of the K-means algorithm with a randomly generated instance where n = 2 ⋅ 106, k = 200, and d = 2.

The models used include VARMA and SVR, whereas the features used include the raw social dynamics (Baseline), level-1 activation vectors (H1), level-2 activation vectors (H2),...

The complete set of matching outcomes and manipulated preferences for each scenario and simulation run can be found in the repository https://zenodo.org/record/2555099 with doi...

The keywords are underlined.

The n values are given in millions.

Time and quality reduction arising from combining F and P algorithms with O-K-means.

Results of O-K-means for a 2-dimensional, non uniform synthetic instance.

Preference overview.

Computer times t and to are given in hours, the product nkd in millions.

Results of OK, FOK and POK when using as initial centroids those generated by K++.