Methods and systems for processing unstructured and unlabelled data

1 . A computer-implemented method, the method comprising:

generating, by a processor, a structured and unlabelled training dataset from an unstructured and unlabelled dataset;

categorizing, by the processor, the structured and unlabelled training dataset into a plurality of clusters by executing an unsupervised learning algorithm, wherein the categorizing includes

clustering the structured and unlabelled training dataset into first clusters with a first k-means clustering algorithm,

determining that first heterogeneous clusters of the first clusters are heterogeneous based on amounts of datapoints of the first heterogeneous clusters, that are related to one or more of first topics or first patterns, being below a threshold,

determining that first homogeneous clusters of the first clusters are homogeneous based on amounts of datapoints of the first homogeneous clusters, that are related to one or more of the first topics or the first patterns, being above the threshold,

clustering the first heterogeneous clusters into second clusters with a second k-means clustering algorithm that is different from the first k-means clustering algorithm,

determining that second heterogeneous clusters of the second clusters are heterogeneous based on amounts of datapoints of the second heterogeneous clusters, that are related to one or more of second topics or second patterns, being below the threshold, and

determining that second homogeneous clusters of the second clusters are homogeneous based on amounts of datapoints of the second homogeneous clusters, that are related to one or more of the second topics or the second patterns, being above the threshold,

wherein each cluster of a selected set of clusters from the plurality of clusters is labelled with an applicable label from a set of labels, wherein the selected set of clusters includes the first and second homogeneous clusters;

training, by the processor, a supervised learning algorithm using a labelled training dataset to generate a trained supervised learning model, the labelled training dataset comprising the set of labels and an input dataset generated from a plurality of datapoints present in each cluster of the selected set of clusters;

generating, by the processor, a Labelled Data1 (LD1) by executing the trained supervised learning model, wherein the trained supervised learning model is configured to assign an applicable label from the set of labels to each datapoint of the first and second heterogeneous clusters;

generating a Labelled Data2 (LD2) by executing the trained supervised learning model, the trained supervised learning model configured to assign an applicable label from the set of labels to each datapoint of a structured and unlabelled retraining dataset;

categorizing the structured and unlabelled retraining dataset into a plurality of unlabelled clusters by executing the unsupervised learning algorithm;

determining homogeneity of the plurality of unlabelled clusters by superimposing the LD2 over the plurality of unlabelled clusters and identifying how many points in each respective clusters of the plurality of unlabelled clusters have a same label of the set of labels; and

retraining the trained supervised learning model based on clusters from the plurality of unlabelled clusters that are determined as being heterogeneous based on the determining homogeneity of the plurality of unlabelled clusters.

2 . The method as claimed in claim 1 , wherein generating the LD1 by executing the trained supervised learning model, further comprises:

assigning an applicable label from the set of labels to each datapoint of a plurality of datapoints present in each cluster of a rejected set of clusters from the plurality of clusters, wherein the rejected set of clusters includes the first and second heterogeneous clusters.

3 . The method as claimed in claim 1 , wherein each cluster of the selected set of clusters is selected based on homogeneity of each cluster of the selected set of clusters.

4 . The method as claimed in claim 1 , wherein the unstructured and unlabelled dataset is one of a textual data, an image data, or a numerical data.

5 . The method as claimed in claim 1 , further comprising:

generating the structured and unlabelled retraining dataset from an unstructured and unlabelled updated dataset.

6 . The method as claimed in claim 1 , further comprising:

identifying a set of heterogeneous unlabelled clusters from the plurality of unlabelled clusters, wherein each heterogeneous unlabelled cluster is labelled with a new label from a set of new labels; and

generating an updated set of labels comprising the set of labels and the set of new labels.

7 . The method as claimed in claim 6 , wherein the retraining includes:

executing the supervised learning algorithm using a labelled retraining dataset to generate a retrained supervised learning model, the labelled retraining dataset comprising the updated set of labels and an updated input dataset generated from a plurality of datapoints present in each unlabelled cluster of the plurality of unlabelled clusters.

8 . The method as claimed in claim 7 , further comprising:

generating an updated labelled data3 (LD3) by executing the retrained supervised learning model, the retrained supervised learning model configured to assign an applicable label from the updated set of labels to each datapoint of the structured and unlabelled retraining dataset.

9 . A computing system comprising:

a memory comprising executable instructions; and

a processor configured to execute the instructions to cause the computing system to at least:

generate a structured and unlabelled training dataset from an unstructured and unlabelled dataset;

categorize the structured and unlabelled training dataset into a plurality of clusters by executing an unsupervised learning algorithm, wherein to categorize the structured and unlabelled training dataset into the plurality of clusters, the processor is configured to execute the instructions to cause the computing system to at least:

cluster the structured and unlabelled training dataset into first clusters with a first k-means clustering algorithm,

determine that first heterogeneous clusters of the first clusters are heterogeneous based on amounts of datapoints of the first heterogeneous clusters, that are related to one or more of first topics or first patterns, being below a threshold,

determine that first homogeneous clusters of the first clusters are homogeneous based on amounts of datapoints of the first homogeneous clusters, that are related to one or more of the first topics or the first patterns, being above the threshold,

cluster the first heterogeneous clusters into second clusters with a second k-means clustering algorithm that is different from the first k-means clustering algorithm,

determine that second heterogeneous clusters of the second clusters are heterogeneous based on amounts of datapoints of the second heterogeneous clusters, that are related to one or more of second topics or second patterns, being below the threshold, and

determine that second homogeneous clusters of the second clusters are homogeneous based on amounts of datapoints of the second homogeneous clusters, that are related to one or more of the second topics or the second patterns, being above the threshold,

wherein each cluster of a selected set of clusters from the plurality of clusters is labelled with an applicable label from a set of labels, wherein the selected set of clusters includes the first and second homogeneous clusters;

train a supervised learning algorithm using a labelled training dataset to generate a trained supervised learning model, the labelled training dataset comprising the set of labels and an input dataset generated from a plurality of datapoints present in each cluster of the selected set of clusters;

generate a Labelled Data1 (LD1) by executing the trained supervised learning model, wherein the trained supervised learning model is configured to assign an applicable label from the set of labels to each datapoint of the first and second heterogeneous clusters;

generate a Labelled Data2 (LD2) by executing the trained supervised learning model, the trained supervised learning model configured to assign an applicable label from the set of labels to each datapoint of a structured and unlabelled retraining dataset;

categorize the structured and unlabelled retraining dataset into a plurality of unlabelled clusters by executing the unsupervised learning algorithm;

determine homogeneity of the plurality of unlabelled clusters by superimposing the LD2 over the plurality of unlabelled clusters and identifying how many points in each respective clusters of the plurality of unlabelled clusters have a same label of the set of labels; and

retrain the trained supervised learning model based on clusters from the plurality of unlabelled clusters that are determined as being heterogeneous based on the determining homogeneity of the plurality of unlabelled clusters.

10 . The computing system as claimed in claim 9 , wherein for generating the LD1 by executing the trained supervised learning model, the computing system is further caused to:

assign an applicable label from the set of labels to each datapoint of a plurality of datapoints present in each cluster of a rejected set of clusters from the plurality of clusters, wherein the rejected set of clusters includes the first and second heterogeneous clusters.

11 . The computing system as claimed in claim 9 , wherein each cluster of the selected set of clusters is selected based on homogeneity of each cluster of the selected set of clusters.

12 . The computing system as claimed in claim 9 , wherein the computing system is further caused to:

generate the structured and unlabelled retraining dataset from an unstructured and unlabelled updated dataset.

13 . The computing system as claimed in claim 9 , wherein the computing system is further caused to:

identify a set of heterogeneous unlabelled clusters from the plurality of unlabelled clusters, wherein each heterogeneous unlabelled cluster is labelled with a new label from a set of new labels; and

generate an updated set of labels comprising the set of labels and the set of new labels.

14 . The computing system as claimed in claim 13 , wherein to perform the retraining, the computing system is further caused to:

execute the supervised learning algorithm using a labelled retraining dataset to generate a retrained supervised learning model, the labelled retraining dataset comprising the updated set of labels and an updated input dataset generated from a plurality of datapoints present in each unlabelled cluster of the plurality of unlabelled clusters.

15 . The computing system as claimed in claim 14 , wherein the computing system is further caused to:

generate an updated labelled data3 (LD3) by executing the retrained supervised learning model, the retrained supervised learning model configured to assign an applicable label from the updated set of labels to each datapoint of the structured and unlabelled retraining dataset.

16 . At least one computer readable storage medium comprising a set of instructions, which when executed by a computing device, cause the computing device to:

generate a structured and unlabelled training dataset from an unstructured and unlabelled dataset;

categorize the structured and unlabelled training dataset into a plurality of clusters by executing an unsupervised learning algorithm, wherein to categorize the structured and unlabelled training dataset into the plurality of clusters, the set of instructions when executed by the computing device, cause the computing device to at least:

cluster the structured and unlabelled training dataset into first clusters with a first k-means clustering algorithm,

determine that first heterogeneous clusters of the first clusters are heterogeneous based on amounts of datapoints of the first heterogeneous clusters, that are related to one or more of first topics or first patterns, being below a threshold,

determine that first homogeneous clusters of the first clusters are homogeneous based on amounts of datapoints of the first homogeneous clusters, that are related to one or more of the first topics or the first patterns, being above the threshold,

cluster the first heterogeneous clusters into second clusters with a second k-means clustering algorithm that is different from the first k-means clustering algorithm,

determine that second heterogeneous clusters of the second clusters are heterogeneous based on amounts of datapoints of the second heterogeneous clusters, that are related to one or more of second topics or second patterns, being below the threshold, and

determine that second homogeneous clusters of the second clusters are homogeneous based on amounts of datapoints of the second homogeneous clusters, that are related to one or more of the second topics or the second patterns, being above the threshold,

wherein each cluster of a selected set of clusters from the plurality of clusters is labelled with an applicable label from a set of labels, wherein the selected set of clusters includes the first and second homogeneous clusters;

train a supervised learning algorithm using a labelled training dataset to generate a trained supervised learning model, the labelled training dataset comprising the set of labels and an input dataset generated from a plurality of datapoints present in each cluster of the selected set of clusters;

generate a Labelled Data1 (LD1) by executing the trained supervised learning model, wherein the trained supervised learning model is configured to assign an applicable label from the set of labels to each datapoint of the first and second heterogeneous clusters;

generate a Labelled Data2 (LD2) by executing the trained supervised learning model, the trained supervised learning model configured to assign an applicable label from the set of labels to each datapoint of a structured and unlabelled retraining dataset;

categorize the structured and unlabelled retraining dataset into a plurality of unlabelled clusters by executing the unsupervised learning algorithm;

determine homogeneity of the plurality of unlabelled clusters by superimposing the LD2 over the plurality of unlabelled clusters and identifying how many points in each respective clusters of the plurality of unlabelled clusters have a same label of the set of labels; and

retrain the trained supervised learning model based on clusters from the plurality of unlabelled clusters that are determined as being heterogeneous based on the determining homogeneity of the plurality of unlabelled clusters.

17 . The at least one computer readable storage medium of claim 16 , wherein the instructions, when executed, cause the computing device to:

assign an applicable label from the set of labels to each datapoint of a plurality of datapoints present in each cluster of a rejected set of clusters from the plurality of clusters, wherein the rejected set of clusters includes the first and second heterogeneous clusters.