| CPC G06F 18/2433 (2023.01) [G06F 17/18 (2013.01); H04N 21/24 (2013.01)] | 10 Claims |
|
1. A system for evaluating multi-dimensional information, comprising:
one or more processors that:
ingest one or more streams of raw, multi-dimensional event data collected via sensors installed on a plurality of remote client devices;
pre-compute session level performance measures at least in part by summarizing the ingested raw, multi-dimensional event data on a per-session basis; and
store the pre-computed session-level performance measures to a data store;
an interface that:
provides a plurality of initial dimensions, wherein each dimension represents a factor related to performance;
receives, from a user, a first selection of a value for a first dimension in the plurality of initial dimensions;
provides a plurality of potentially significant dimensions from among a set of dimensions, wherein the set of dimensions comprises a search space determined based at least in part on the first selection of the value for the first dimension; and
receives a second selection of a second dimension from among the plurality of potentially significant dimensions; and
wherein the one or more processors further:
determine the plurality of potentially significant dimensions based on an indication of presence of outliers in the plurality of potentially significant dimensions, wherein a dimension that significantly affects performance is identified, and wherein the dimension that significantly affects performance is identified at least in part by scanning, at query time, the search space at least in part by:
determining a set of performance measures at least in part by accessing the data store of precomputed session level performance measures based at least in part on the first selection of the value by the user;
grouping the set of performance measures into cohorts of performance measures, wherein each cohort corresponds to a dimension value of the dimension, and wherein the dimension comprises a plurality of dimension values;
generating, for each cohort, a corresponding cohort-level performance measure at least in part by dynamically performing, at query time, an aggregation across pre-computed session-level performance measures belonging to a given cohort;
fitting, at query time, a beta distribution to the cohort-level performance measures; and
detecting a presence of an outlier in the dimension based at least in part on a parameter associated with the beta distribution fitted, at query time, to the cohort-level performance measures, wherein the detected presence of the outlier in the dimension is indicative of anomalous performance behavior; and based at least in part on the scanning, at query time, of the search space, dynamically update the interface, including:
surfacing, as a selectable option, the identified dimension that significantly affects performance to the user as a recommended next dimension to explore, wherein the second selection is received in response to the user clicking on the identified dimension surfaced in the interface as the selectable option, the second dimension comprising the identified dimension; and
responsive to the user clicking on the identified dimension surfaced in the interface, updating a panel in the interface, including visually indicating ratings of performance of sessions grouped according to different values of the identified dimension clicked on by the user, wherein the ratings of performance are determined according to a multi-level scale.
|