PARENTAL COMMUNITIES CLASSIFICATION
Disclosed are methods for predicting assignments of users to groups based on obtained data. The methods may include phasing the obtained data of a user and inputting features of the phased data to a trained model to receive an output that provides information associated with a group assignment of the user. The model may be trained with a plurality of training samples, each training sample comprising features of phased user data of a reference user and a label identifying whether the reference user belongs to a group. Training the model may include applying a plurality of classifiers to at least one of the training samples to obtain an output from each classifier, determining an average of the outputs from applying the plurality of classifiers, and determining the average of the plurality of classifiers as the output from the model.
1 . A computer-implemented method, comprising:
phasing user data associated with a named entity to generate a pair of datasets of the named entity;
selecting features of one of the datasets of the pair of datasets of the named entity;
inputting the selected features into a trained model, wherein the model has been trained by:
obtaining a plurality of training samples, each training sample comprising features of a reference named entity and a label identifying whether the reference named entity belongs to a community, wherein reference named entities who belong to the community are representatives of the community, and
applying the model to the plurality of training samples;
receiving an output from the model, the output providing information associated with a community assignment of the named entity.
2 . The method of claim 1 , wherein applying the model to the plurality of training samples comprises:
applying a plurality of classifiers to at least one of the training samples to obtain an output from each classifier, comprising:
for at least one of the plurality of classifiers,
randomly selecting at least one of the training samples using Bagging; and
applying the at least one of the plurality of classifiers to the selected at least one training sample to obtain an output from the at least one of the plurality of classifiers.
3 . The method of claim 2 , wherein at least one of the plurality of classifiers is XGBoost.
4 . The method of claim 1 , wherein obtaining a plurality of training samples comprises:
receiving target user data from a target named entity, wherein the target named entity is associated with a first label identifying whether the target named entity belongs to a first community;
phasing the target user data of the target named entity to generate a pair of target datasets of the target named entity;
comparing each of the target datasets of the target named entity to a dataset of a reference named entity who belongs to the first community;
determining a number of matches with the dataset of the reference named entity between each of the target datasets of the target named entity; and
determining the target dataset having a higher number of matches as the selected dataset of phased user data of the target named entity in a training sample.
5 . The method of claim 1 , wherein the output from the model includes a prediction that the named entity belongs to a community and a confidence level of the community assignment.
6 . The method of claim 1 , wherein training the model further comprises:
selecting a uniquely common dataset as the selected dataset of the phased user data that is representative of the community, wherein the uniquely common dataset is most likely to be observed in the community.
7 . The method of claim 6 , wherein training the model further comprises:
generating a feature vector for at least one of the training samples for the community based on the selected dataset of the phased user data, the feature vector comprising one or more non-zero feature elements.
8 . A computer-readable storage medium having stored thereon computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations, comprising:
phasing user data associated with a named entity to generate a pair of datasets of the named entity;
selecting features of one of the datasets of the pair of datasets of the named entity;
inputting the selected features into a trained model, wherein the model has been trained by:
obtaining a plurality of training samples, each training sample comprising a selected dataset of phased user data of a reference named entity and a label identifying whether the reference named entity belongs to a community, wherein reference named entities who belong to the community are representative of the community, and
applying the model to the plurality of training samples; and
receiving an output from the model, the output providing information associated with a community assignment of the named entity.
9 . The computer-readable storage medium of claim 8 , wherein applying the model to the plurality of training samples comprises:
applying a plurality of classifiers to at least one of the training samples to obtain an output from each classifier, comprising:
for at least one of the plurality of classifiers,
randomly selecting at least one of the training samples using Bagging; and
applying the at least one of the plurality of classifiers to the selected at least one training sample to obtain an output from the at least one of the plurality of classifiers.
10 . The computer-readable storage medium of claim 9 , wherein at least one of the plurality of classifiers is XGBoost.
11 . The computer-readable storage medium of claim 8 , wherein obtaining a plurality of training samples comprises:
receiving target user data from a target named entity, wherein the target named entity is associated with a first label identifying whether the target named entity belongs to a first community;
phasing the target user data of the target named entity to generate a pair of target datasets of the target named entity;
comparing each of the target datasets of the target named entity to a dataset of a reference named entity who belongs to the first community;
determining a number of matches with the dataset of the reference named entity between each of the target datasets of the target named entity; and
determining the target dataset having a higher number of matches as the selected dataset of phased user data of the target named entity in a training sample.
12 . The computer-readable storage medium of claim 8 , wherein the output from the model includes a prediction that the named entity belongs to a community and a confidence level of the community assignment.
13 . The computer-readable storage medium of claim 8 , wherein training the model further comprises:
selecting a uniquely common dataset as the selected dataset of the phased user data that is representative of the community, wherein the uniquely common dataset is most likely to be observed in the community.
14 . The computer-readable storage medium of claim 13 , wherein training the model further comprises:
generating a feature vector for at least one of the training samples for the community based on the selected dataset of the phased user data, the feature vector comprising one or more non-zero feature elements.
15 . A computer system, comprising:
one or more processors; and
a hardware storage device having stored thereon computer-executable instructions that, when executed by the one or more processors, causes the computer system to perform operations, comprising:
phasing user data associated with a named entity to generate a pair of datasets of the named entity;
selecting features of one of the datasets of the pair of datasets of the named entity;
inputting the selected features into a trained model, wherein the model has been trained by:
obtaining a plurality of training samples, each training sample comprising a selected dataset of phased user data of a reference named entity and a label identifying whether the reference named entity belongs to a community, wherein reference named entities who belong to the community are representative of the community, and
applying the model to the plurality of training samples; and
receiving an output from the model, the output providing information associated with a community assignment of the named entity.
16 . The computer system of claim 15 , wherein applying the model to the plurality of training samples comprises:
applying a plurality of classifiers to at least one of the training samples to obtain an output from each classifier, comprising:
for at least one of the plurality of classifiers,
randomly selecting at least one of the training samples using Bagging; and
applying the at least one of the plurality of classifiers to the selected at least one training sample to obtain an output from the at least one of the plurality of classifiers.
17 . The computer system of claim 16 , wherein at least one of the plurality of classifiers is XGBoost.
18 . The computer system of claim 15 , wherein obtaining a plurality of training samples comprises:
receiving target user data from a target named entity, wherein the target named entity is associated with a first label identifying whether the target named entity belongs to a first community;
phasing the target user data of the target named entity to generate a pair of target datasets of the target named entity;
comparing each of the target datasets of the target named entity to a dataset of a reference named entity who belongs to the first community;
determining a number of matches with the dataset of the reference named entity between each of the target datasets of the target named entity; and
determining the target dataset having a higher number of matches as the selected dataset of phased user data of the target named entity in a training sample.
19 . The computer system of claim 15 , wherein the output from the model includes a prediction that the named entity belongs to a community and a confidence level of the community assignment.
20 . The computer system of claim 15 , wherein training the model further comprises:
selecting a uniquely common dataset as the selected dataset of the phased user data that is representative of the community, wherein the uniquely common dataset is most likely to be observed in the community.