Method and apparatus for customer experience segmentation based on a web session event variation
View Patent ↗Web session events are captured during different web sessions. A segmentation model is generated based on a number of occurrences of the web session events for different time stamp periods. The segmentation model plots a segmentation graph that may identify differences between the web session events with respect to time factors. The segmentation model may use the whole dataset of event occurrences as an input without any preliminary data segmentation or discrimination. The model can associate the web session events with the different geographic locations, reveal possible reasons for customer experience difference for the users from different locations and provide statistically sound explanation of this difference. The model is scalable and may work with big data acquired by web-based commerce sites with wide international customer base.
1. A computer-implemented method, comprising:
identifying events for web sessions;
identifying a number of occurrences of the events for different times during a day;
generating a segmentation model from the number of occurrences of the events for the different times during the day;
using the segmentation model to produce a scatter plot having a first set of data points associated with a maximum number of occurrences for each of the events and a second set of data points associated with time factors identifying times of the day for the maximum number of occurrences for each of the events;
segmenting the events into different geographic regions based on relationships between the first set of data points associated with the maximum number of occurrences for each of the events and the second set of data points associated with the time factors; and
using the segmentation model to find clusters of the events that are correlated in the scatter plot, wherein the relationships between the first set of data points and the second set of data points are identified by comparing positions of the first set of data points with directions, as defined by the second set of data points, from an origin point of a cluster in the scatter plot.
2. The method of claim 1 , wherein the segmentation model projects the first and second set of data points on the scatter plot in a circle around the origin point, and the first set of data points are associated with the second set of data points aligned in a same radial direction on the scatter plot.
3. The method of claim 2 , wherein a distance of the first set of data points from the origin point indicates an amplitude of oscillation for the number of occurrences of the events for the web sessions.
4. The method of claim 1 , wherein segmenting the events comprises identifying groups of the data points associated with same times of the day.
5. The method of claim 1 , wherein segmenting the events into different geographic regions comprises identifying groups of the events with the largest number of occurrences during a particular time period during the day and associating the groups of the events with the different geographic regions.
6. The method of claim 1 , wherein segmenting the events into different geographic regions comprises associating the different geographic regions with different sectors of the scatter plot based on the directions.
7. The method of claim 1 , wherein the segmentation model identifies oscillation parameters for a periodicity of the occurrences of the events.
8. The method of claim 1 , wherein the events comprise interactions between a web application and users during the web sessions.
9. The method of claim 1 , wherein the segmentation model comprises a partial least squares discriminant analysis (PLS-DA) model.
10. An apparatus, comprising:
a memory configured to archive web session events for web sessions; and
logic circuitry configured to:
extract the web session events from the memory;
generate a segmentation model comprising segmentation factors for the web session events based on a number of occurrences of the web session events for different time periods, wherein the segmentation factors associate different geographic locations to the different time periods based on an expected peak in the number of occurrences of the web session events for the different time periods;
identify the different geographic locations for the web session events based on relationships between a first set of data points associated with the number of occurrences for the web session events and a second set of data points associated with the segmentation factors; and
using the segmentation model to find clusters of the web session events that are correlated in a scatter plot produced by the segmentation model, wherein the relationships between the first set of data points and the second set of data points are identified by comparing positions of the first set of data points with directions, as defined by the second set of data points, from an origin point of a cluster in the scatter plot.
11. The apparatus of claim 10 , wherein the segmentation factors are associated with the different time periods.
12. The apparatus of claim 10 , wherein the segmentation factors comprise document object model (DOM) events.
13. The apparatus of claim 10 , wherein the logic circuitry is further configured to use a partial least squares discriminant analysis (PLS-DA) model to identify the different geographic locations for the web session events relative to the segmentation factors.
14. The apparatus of claim 10 , wherein each of the segmentation factors correspond to one of the different geographic locations.
15. A computer-implemented method, comprising:
identifying web session events comprising interactions between a web server and computing devices during web sessions;
aggregating occurrences of the web session events for timestamp periods;
generating a segmentation model based on the occurrences of the web session events for the timestamp periods;
using the segmentation model to describe the web session events as a first set of data points in a partial least squares (PLS) components space associated with the occurrences of the web session events for the timestamp periods;
using the segmentation model to describe segmentation factors as a second set of data points in the PLS components space, wherein the segmentation factors associate different geographic locations to the timestamp periods based on an expected peak in the occurrences of the web session events for the timestamp periods;
identifying the different geographic locations for the web session events based on relationships between the first set of data points and the second set of data points; and
using the segmentation model to find clusters of the web session events that are correlated in the PLS components space, wherein the relationships between the first set of data points and the second set of data points are identified by comparing positions of the first set of data points with directions, as defined by the second set of data points, from an origin point of a cluster in the PLS components space.
16. The method of claim 15 , wherein the segmentation model is used to find the web session events that are correlated in the PLS components space by detecting data points associated with the clusters of web events located in same sectors of the PLS components space.