Leverage interaction attributes and file access patterns to predict active life of media file to optimize storage usage

1. A computerized-method for determining a storage class of a media file, said computerized-method comprising:

in a computerized system comprising a processor, one or more databases of a plurality of data sources, and a memory to store the one or more databases, said processor is configured to:

(i) aggregating data from the plurality of data sources;

(ii) operating a pre-built machine learning File Access Frequency Score (FAFS) model based on the aggregated data to calculate an accessibility score for a media file;

(iii) sending the media file to archive storage when the accessibility score is below or equals a preconfigured threshold;

(iv) calculating Predicted Active Lifetime Span (PALS) score of the media file, when the accessibility score is above the preconfigured threshold and when the accessibility score is below or equals the preconfigured threshold the PALS score equals zero;

(v) repeating the calculating of the PALS score of the media file every preconfigured period of time, until the calculated PALS score of the media the is lower than or equals zero; and

(vi) sending the media file to the archive storage,

wherein the media file is a recording of an interaction and wherein the aggregated data is related to the interaction and to a customer and an agent participating in the interaction.

2. The computerized-method of claim 1 , wherein the storage class has related accessibility level and related cost, and wherein the storage class is selected from at least one of: archive storage and active storage.

3. The computerized-method of claim 1 , wherein the pre-built machine learning FAFS model is created by:

(i) collecting data from data sources, wherein the data is retrieved from related historical data databases, and wherein labeling of the data is performed according to a first access to the media file by a user;

(ii) preprocessing the collected data;

(iii) training the machine learning FAFS model; and

(iv) performing model evaluation and algorithm selection.

4. The computerized-method of claim 1 , wherein the PALS score is calculated by a weighted sum of: (i) a preconfigured constant; (ii) FAFS score; (iii) a number of file access frequency in active storage; and (iv) time elapsed since the media file has been last accessed.

5. The computerized-method of claim 4 , wherein the preconfigured constant is determined according to a time period that a media file is required to be accessible.

6. The computerized-method of claim 4 , wherein the weights are determined according to a time period that a media file is required to be accessible.

7. The computerized-method of claim 1 , wherein the data sources are selected from at least one of: recording applications, omnichannel interaction recording, email channel, chat channel, storage services, analytics application, and quality management module.

8. The computerized-method of claim 1 , wherein the aggregated data from the data sources that is related to customers is selected from at least one of: customer feedback received for the interaction, customer sentiment score on the interaction, pattern of customer feedback, and an average feedback rating provided by customers in a given interaction context.

9. The computerized-method of claim 1 , wherein the aggregated data from the data sources that is related to agents is selected from at least one of:

an agent historical performance for a skill used in the interaction, a skill used to route a contact to available agents according to customer's needs and agent feedback about the customer and a team the agent belongs to.

10. The computerized-method of claim 1 , wherein the aggregated data from the data sources that is related to interactions is selected from at least one of:

duration of the interaction as compared to Service Level Agreement (SLA), type of file, a campaign under which an interaction has been taken, and an average active storage time span of files received in similar context across customers in same domain.