AUTOMATED ENTITY CLASSIFICATION USING USAGE HISTOGRAMS & ENSEMBLES
Techniques disclosed herein employ entity-activity data expressed in a discrete distribution (histogram) form having one or many dimensions to dynamically classify the entity's usage and/or behavior patterns, where groupings or segmentations of different entities that exhibit similar usage patterns are identified using various approaches, including dimensionality reduction, and/or clustering procedures. A consensus or ensemble clustering may be generated that represents a clustering of clusters, based on subclusterings themselves, and/or any combination of subclusters with entity-activity data to selectively execute a market offering campaign. In one embodiment, the resulting ensemble clusterings enable selective directing of targeted offerings to a telecommunication provider's customers.
1 . A network device, comprising:
a transceiver to send and receive data over a network; and
a processor that performs actions, comprising:
receiving telecommunications customer data for a plurality of customers;
extracting from the customer data a usage histogram for each of the plurality of customers;
computing for each of the plurality of customers, a reduced dimensionality usage histogram from the extracted usage histograms;
performing a clustering from the reduced dimensionality usage histograms to generate a plurality of clusters; and
classifying each customer time series within one of the plurality of clusters, the classifications selectively usable to dynamically market to a customer identified by a cluster.
2 . The network device of claim 1 , wherein for each of the plurality of customers the processor performs actions, further comprising:
combining the cluster classification from the usage histogram content with cluster classifications from other clustering solutions;
performing a consensus clustering of the combined cluster classifications;
classifying each customer with the consensus cluster assignment, the classifications usable to dynamically market to at least one customer identified by the consensus cluster.
3 . The network device of claim 2 , wherein combining the cluster classifications further comprises combining the cluster classifications with at least some of the received telecommunications customer data prior to performing the consensus clustering.
4 . The network device of claim 1 , wherein the clusters being selectively usable to dynamically market to a customer identified by a cluster, further comprises:
employing a threshold value that is applied to data within a cluster to determine whether to provide an offering at a given time or location to a given customer; and
when it is determined that the offering has a likelihood of not being accepted by the given customer based on the threshold for the given time and location, then inhibiting sending of the offering to the given customer; and otherwise,
sending the offering to the given customer at the given time or location.
5 . The network device of claim 1 , wherein performing a clustering from the reduced dimensionality usage histogram to generate a plurality of clusters, further comprising:
determining a number of clusters to generate in the plurality of clusters using a statistical measure of an orthogonality of data types within the telecommunications customer data for the plurality of customers.
6 . The network device of claim 1 , wherein the usage histograms are represented using matrix-factorized histogram coefficients.
7 . The network device of claim 1 , wherein classifying each customer time series is based on training of a behavioral classification model that employs a cross-validation mechanism to select a minimum number of training patterns to satisfy a selected criteria.
8 . The network device of claim 1 , wherein for each of the plurality of customers the processor performs actions, further comprising:
combining the cluster classification from the usage histogram content with cluster classifications from a defined number of other clustering solutions, the number of clusters that are combined is determined based on a number of basis vectors obtained in a non-negative matrix factorization decomposition of a training set of data.
9 . A system, comprising:
one or more non-transitory storage devices usable to store customer data; and
one or more processors that perform actions, comprising:
receiving telecommunications customer data for a plurality of customers;
extracting from the telecommunications customer data a usage histogram for each of the plurality of customers, wherein each histogram includes a customer's usage pattern over a given time window;
computing for each of the plurality of customers, a reduced dimensionality usage histogram from the extracted usage histograms;
performing a clustering from the reduced dimensionality usage histogram to generate a plurality of clusters; and
classifying each customer time series within one of the plurality of clusters, the classifications selectively used to dynamically identify an occasion when to perform an interaction directed towards a customer identified by a cluster.
10 . The system of claim 9 , wherein computing for each of the plurality of customers, a reduced dimensionality usage histogram includes using a non-negative matrix factorization to generate a number of basis vectors.
11 . The system of claim 9 , wherein classifying each customer time series is based on training of a behavioral classification model that employs a cross-validation mechanism to select a minimum number of training patterns to satisfy a selected criteria.
12 . The system of claim 9 , wherein for each of the plurality of customers the one or more processors perform actions, further comprising:
combining the cluster classification from the usage histogram content with cluster classifications from other clustering solutions;
performing a consensus clustering of the combined cluster classifications;
classifying each customer within the consensus cluster assignment, the classifications being used to dynamically market to at least one customer identified by a consensus cluster.
13 . The system of claim 12 , wherein combining the cluster classifications further comprises combining the cluster classifications with at least some of the received telecommunications customer data prior to performing the consensus clustering.
14 . The system of claim 9 , wherein at least one of the other clustering solutions is determined using a different clustering technique than that used for determining the cluster classification.
15 . The system of claim 9 , wherein the clusters being selectively used to dynamically market to a customer identified by a cluster, further comprises:
employing a threshold value that is applied to data within a cluster to determine whether to provide an offering at a given time or location to a given customer; and
when it is determined that the offering has a likelihood of not being accepted by the given customer based on the threshold for the given time and location, then inhibiting sending of the offering to the given customer; and otherwise,
sending the offering to the given customer at the given time or location.
16 . The system of claim 9 , wherein performing a clustering from the reduced dimensionality usage histogram to generate a plurality of clusters, further comprising:
determining a number of clusters to generate in the plurality of clusters using a statistical measure of an orthogonality of data types within the telecommunications customer data for the plurality of customers.
17 . An apparatus comprising a non-transitory computer readable medium, having computer-executable instructions stored thereon, that in response to execution by a special purpose computing device, cause the special purpose computing device to perform operations, comprising:
receiving telecommunications customer data for a plurality of customers;
extracting from the telecommunications customer data a usage histogram for each of the plurality of customers, wherein each histogram includes a customer's usage pattern over a given time window;
computing for each of the plurality of customers, a reduced dimensionality usage histogram from the extracted usage histograms;
performing a clustering from the reduced dimensionality usage histogram to generate a plurality of clusters; and
classifying each customer time series within one of the plurality of clusters, the classifications selectively being used to dynamically identify an occasion when to perform an interaction directed towards a customer identified by a cluster.
18 . The apparatus of claim 17 , wherein for each of the plurality of customers the special purpose computing device to perform operations, further comprising:
combining the cluster classification from the usage histogram content with cluster classifications from other clustering solutions;
performing a consensus clustering of the combined cluster classifications;
classifying each customer with the consensus cluster assignment, the classifications selectively used to dynamically market to at least one customer identified by a consensus cluster.
19 . The apparatus of claim 18 , wherein combining the cluster classifications further comprises combining the cluster classifications with at least some of the received telecommunications customer data.
20 . The apparatus of claim 17 , wherein the clusters being selectively used to dynamically market to a customer identified by a cluster, further comprises:
employing a threshold value that is applied to data within a cluster to determine whether to provide an offering at a given time or location to a given customer; and
when it is determined that the offering has a likelihood of not being accepted by the given customer based on the threshold for the given time and location, then inhibiting sending of the offering to the given customer; and otherwise,
sending the offering to the given customer at the given time or location.
21 . The apparatus of claim 17 , wherein performing a clustering from the reduced dimensionality usage histogram to generate a plurality of clusters, further comprising:
determining a number of clusters to generate in the plurality of clusters using a statistical measure of an orthogonality of data types within the telecommunications customer data for the plurality of customers.