SYSTEM AND METHOD FOR AUTOMATED CHRONIC ISSUE RESOLUTION IN CUSTOMER PREMISES EQUIPMENT
BACKGROUND
[0001] Embodiments of the present specification relate generally to data analytics, and more particularly to a system and method for automated critical problem resolution in customer premises equipment based on progressive analytic modelling.
[0002] Consumer demand for media-rich home entertainment services is driving innovation and expansion of the services delivered to users via customer premises equipment (CPE). Particularly, present day CPEs are increasingly implemented as hybrid devices, integrating video content from multiple signal sources such as broadcast television, video-on-demand, and internet-based interactive games and services. Additionally, the CPE devices may also provide value-added capabilities like time-shifting, user-initiated re-configuration, and allowing content to be distributed to multiple devices such as televisions, personal computers, portable media players, and other mobile devices.
[0003] With the advent of such intelligent CPE devices, service providers strive to find efficient ways for managing these devices and associated functionality to ensure provision of superior services without undue disruptions. However, failures in the CPE, communications network, or reduced quality of service (QoS) in the reception of information significantly degrade the quality of user experience (QoE). Conventionally, a user experiencing disruption contacts a customer service or repair center to troubleshoot the set top box. Typical resolution of user complaints entails sending troubleshooting signals such as a reboot signal to the CPE, and/or sending customer service technician to the user premises to rectify the error.

[0004] Troubleshooting customer premises equipment over phone and/or in person, however, involves significant time and expense for the service provider. Moreover, such black-box testing methodology provides only limited information to identify root cause of a reported issue. Certain service providers, therefore, employ CPE devices that allow remote access to device information stored in logs for manual and/or automatic analysis for identifying a root cause of the reported issue.
[0005] However, historically, some chronic issues occur at random for certain rare combination of operational parameters, usage patterns, and/or resulting devices states. Such chronic issues may not be reproducible in lab even with the help of the remotely accessible log information due to the significantly large number of operational parameters, usage patterns, and/or resulting devices states. Furthermore, due to the rare and random nature of the issue, use of conventional data analytics may prove infeasible as it would involve significant number of trial and error iterations for identifying a resolution via theory of elimination.
[0006] Failure to reproduce the issue hinders the ability to identify the root cause, and in turn, an acceptable solution to the chronic issue. Such issues, thus, often continue to be chronic or known problems with an unknown cause leading to investment of substantial time and money, while hampering customer goodwill and/or brand image of the service provider. Accordingly, it may be desirable to find methods or systems that quickly and accurately identify root cause of chronic issues associated with CPE without undue disruption to the service provided to the users.
BRIEF DESCRIPTION
[0007] In accordance with certain aspects of the present specification, a method for critical problem resolution is presented. The method includes processing data corresponding to one or more connected devices operable in a communications

network to define one or more boundary conditions corresponding to the chronic issue associated with a connected device in the plurality of connected devices. The method also includes identifying one or more candidate parameters from the processed data that are determined to cause greatest variability of the data. The method further includes determining one or more issue predictor models that aggregate a suite of learning algorithms to evaluate effect of one or more of the identified candidate parameters on occurrence of the chronic issue. Additionally, the method includes iteratively developing an outcome model from one or more of the issue predictor models based on a statistical analysis of performance of each of the issue predictor models such that the outcome model satisfies one or more specified exit criteria so as to identify occurrence of a root cause of the chronic issue with a desired accuracy level. Moreover, the method includes communicating the root cause of the chronic issue to an associated test automation system.
[0008] In accordance with certain further aspects of the present specification, a system for critical problem resolution is presented. The system includes a storage subsystem configured to store data corresponding to one or more connected devices operable in a communications network. The system further includes an analytics subsystem communicatively coupled to the storage subsystem and configured to process the data to define one or more boundary conditions corresponding to the chronic issue associated with a connected device in the plurality of connected devices, identify one or more candidate parameters from the processed data that are determined to cause greatest variability of the data, determine one or more issue predictor models that aggregate a suite of learning algorithms to evaluate effect of one or more of the identified candidate parameters on occurrence of the chronic issue, and iteratively develop an outcome model from one or more of the issue predictor models based on a statistical analysis of performance of each of the issue predictor models such that the outcome model satisfies one or more specified exit criteria so as to identify occurrence of a root cause of the chronic issue with a desired accuracy level. The system also

includes a test automation system communicatively coupled to one or more of the storage subsystem and the analytics subsystem, and configured to receive the identified root cause of the chronic issue from one or more of the storage subsystem and the analytics subsystem, and automatically select one or more test vectors, test scripts, test cases, or combinations thereof, from a repository based on the root cause of the chronic issue to accurately reproduce the chronic issue.
[0009] In accordance with certain other aspects of the present specification a system and method for critical problem resolution in connected devices is disclosed. Particularly, the system and method provide a highly accurate problem prediction model that is developed using a collection of powerful statistical machine learning algorithms. This model, when applied on a data set retrieved in view of the chronic issue, identifies a selected set of parameters and their weightages directly responsible for causing the chronic issue. Furthermore, using the model parameters and corresponding parametric representation, test package can be generated for helping with problem resolution. Use of the model-based approach in highly automated scenario reduces substantial testing effort and expense, while allowing for efficient isolation of the chronic issue, test case selection, and subsequent resolution.
DRAWINGS
[0010] These and other features, aspects, and advantages of the claimed subject matter will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
[0011] FIG. 1 is a schematic representation of an exemplary test system for providing automated chronic issue resolution in CPE;

[0012] FIG. 2 is a graphical representation of an exemplary chronic issue that was automatically identified using the system of FIG. 1;
[0013] FIG. 3 a graphical representation of another exemplary chronic issue that was automatically identified using the system of FIG. 1;
[0014] FIG. 4 is a schematic representation of an exemplary process flow for providing automated chronic issue identification in CPE using the system of FIG. 1;
[0015] FIG. 5 is an exemplary diagram depicting a pattern of parameters identified using the system of FIG. 1 that cause a particular chronic issue in a connected device to aids in selection of suitable test cases for faster resolution of the issue; and
[0016] FIG. 6 is a graphical representation of statistics depicting a comparison of customer-reports of a chronic issue and instances of the same chronic issue predicted using the system of FIG. 1.
DETAILED DESCRIPTION
[0017] The following description presents exemplary systems and methods for automated chronic issue resolution in connected devices such as CPE. Particularly, embodiments described herein disclose systems and methods that aid in accurately identifying root cause of a chronic issue associated with a CPE and a corresponding resolution. As used herein, the terms "chronic issue" and/or "chronic problem" may be used to refer to an issue associated with a CPE that occurs rarely and/or randomly, for example once every six months, and which may not be reproduced in a testing scenario using conventional testing and/or log-based information. The chronic issue, for example, may include an occasional shutdown, unexpected reboot, failure to respond to a key press or channel selection command, data loss, poor signal quality, and/or other related issues.

[0018] For a single CPE used by a particular customer, an issue may be rare and easily ignored. Such issues may be often under-reported and difficult to reproduce, thus impeding identification of a suitable approach for resolution. However, evaluation of data compiled from a few million devices may indicate a significant number of occurrences of a chronic issue. In a sample study of set top box (STB) devices, it was found that long-standing chronic issues constituted from about 4 % to about 5% of total issues, and played a significant role in hampering customer satisfaction and product experience. An exemplary framework that is suitable for practicing various implementations of the present system for automated identification of such chronic issues in a CPE is discussed in the following sections with reference to FIG. 1.
[0019] FIG. 1 illustrates an exemplary system 100 for providing automated chronic issue resolution in a connected device such as a CPE 101. For clarity, the present embodiment is described with reference to automated testing of an STB device (hereinafter referred to as STB 101). However, it may be noted that embodiments of the system 100 may be used for automated chronic issue identification and/or resolution for other connected devices. The connected devices, for example, may include a digital video recorder (DVR), a digital subscriber line (DSL) modem, a broadband router, a network gateway device, a television, voice over internet protocol (VOIP) phone, a desktop computer, a mobile device, and/or an appliance connected to a communications network. For example, the system 100 may be used to identify and resolve an unexpected reset or reboot of a DSL modem, or absence of dial tone of a phone for a short period of time that may or may not get reported by a user.
[0020] In one embodiment, the STB 101 is configured to decode signals received from a service provider, for example, via cable or satellite links and display a resulting audio and/or video (AV) feed on a display device such as a television screen (not shown). In the event of a potential or reported chronic issue with respect to the STB

101, the system 100 employs a progressive analytic model to identify a root cause of the chronic issue for resolution. To that end, the system 100 uses existing customer and/or device-related information stored in a plurality of associated databases. The databases, for example, may include a hard disk drive, a floppy disk drive, a compact disk-read/write (CD-R/W) drive, a Digital Versatile Disc (DVD) drive, a flash drive, and/or a solid state storage device.
[0021] Particularly, in one embodiment, the system 100 retrieves a customer identifier, a device identifier, and/or logs of operational parameters of the STB 101 and/or other STBs in the same communications network (not shown) for a selected period of time, for example, from an associated TR 069 database 102. The TR 069 database 102 may adhere to technical specifications defined in the Technical Report 069 report by the Broadband Forum. Further, the system 100 retrieves desired data logs related to a current and/or historical occurrence of the chronic issue in the STB 101 and/or other STBs on the network, other issues associated with the STB 101, and/or any corresponding resolutions previously provided to the customer from a contact center database 104 and/or a field database 106. Additionally, the system 100 may also retrieve customer management information including location, billing, usage, and/or demographic information from a customer management system 108. Moreover, the system 100 may also be communicatively coupled to a cloud hosting system 109 for access to other data related to the STB 101.
[0022] In certain embodiments, the system 100 further includes a data integration framework 110 that is configured to process the retrieved data to suitably define a chronic issue, boundary conditions corresponding to the chronic issue, and/or components of the service provision ecosystem that are potentially involved in the occurrence of the chronic issue. In certain embodiments, the data integration framework 110 may also use relevant information received from a domain expert that may be used by the system 100 for accurately defining the chronic issue. Further, the

data integration framework 110 also aggregates and/or sanitizes logs, records, and/or transactions for a large number of STB devices connected over the communications network irrespective of whether they have encountered the chronic issue. Subsequently, the data integration framework 110 communicates the aggregated and/or processed data to an analytics subsystem 112 for further processing.
[0023] Accordingly, in one embodiment, the analytics subsystem 112 may include artificial intelligence clustering subsystem 114, a data correlation engine 116, and a predictor identifier 118, which in operative association with each other, process the received data to identify optimal predictors. To that end, the analytics subsystem 112, the artificial intelligence clustering subsystem 114, the data correlation engine 116, and the predictor identifier 118 may be implemented using a one or more application-specific processors, graphical processing units (GPUs), digital signal processors (DSPs), microcomputers, microcontrollers, Application Specific Integrated Circuits (ASICs) and/or Field Programmable Gate Arrays (FPGAs), and/or suitable processing and/or storage devices.
[0024] According to certain aspects of the present specification, the analytics subsystem 112, and components thereof, employ selected data preprocessing techniques to identify one or more parameters that may have direct and large influence on the occurrence of the chronic issue. In one embodiment, the data from the data integration framework 110 may include thousands of candidate parameters that are indicative of different operational characteristics, usage patterns, and/or states of the different STB devices and are recorded by different components in the communications network.
[0025] Specifically, in one embodiment, the analytics subsystem 112 employs a plurality of data analytics algorithms for determining a number of levels, determining a suitable probability distribution function for each candidate parameter, for identifying connections between candidate parameters, and/or combinations of

candidate parameters causing a maximum variability of the recorded data. In certain embodiments, the analytics subsystem 112 may employ the algorithms for stochastic modeling, and/or for determining various transform operators to handle data diversity, centrality, and/or scaling of the received data during evaluation of the different parameters and/or parameter combinations. Additionally, the analytics subsystem 112 may also employ the algorithms for generating a loss function and evaluating each parameter and/or parameter combination against the loss function to identify the most relevant parameters and/or parameter combinations.
[0026] Thus, a resulting output of the analytics subsystem 112 identifies a subset of parameters and/or parameter combinations and their corresponding weightage that have been determined to impact the occurrence of the chronic issue as predictors. Further, the analytics subsystem 112 communicates the list of identified predictors to a predictor modeling subsystem 120. In certain embodiments, the predictor modeling subsystem 120 is configured to build one or more issue predictor models 122 that aggregate a suite of deep learning algorithms and iteratively develop the optimal issue predictor model 122 that best explains the occurrence of the chronic issue.
[0027] Particularly, in one embodiment, an appropriate suite of algorithms is selected based on the type of chronic issue and type of predictors identified by the analytics subsystem 112 to be of specific relevance to the occurrence of the chronic issue. The suite of algorithms, for example, may include support vector machines, classification, pattern recognition, logistic regression, linear discriminant analysis, decision tree, neural network, and/or Bayesian algorithms that build the one or more issue predictor models 122 based on the identified predictors. Further, the predictor modeling subsystem 120 defines an objective function for cost optimization corresponding to each of the issue predictor models 122. Additionally, the predictor modeling subsystem 120 selects a set of control parameters or feature vectors for the issue predictor model 122 that provide the best model performance. Particularly, the

predictor modeling subsystem 120 selects performance evaluated/accurately modeled parameters, which explain a variance in observable data. In one embodiment, the selection of the optimal parameters is performed either in feature space or component space. P-value is computed to determine relevance of the parameter to the chronic issue. By way of example, the lower the p-value, the more accurate the feature or parameter is for explaining the variance in the output.
[0028] Further, in certain embodiments, the predictor modeling subsystem 120 performs a statistical analysis of the performance of each of the issue predictor models 122 to iteratively identify an outcome model generated from a single or a suitable combination of the one or more issue predictor models 122. Particularly, the outcome model may be iteratively identified once one or more specified exit criteria for achieving desired model performance is reached. The resulting outcome model that best explains the chronic issue may be used for establishing the identified root cause of the problem.
[0029] Subsequently, the predictor modeling subsystem 120 communicates the information identifying the root cause of the problem to a test automation system (TAS) 124. Based on the root cause identified by the outcome model, the TAS 124 identifies appropriate test vectors or test cases for accurate issue reproduction and resolution. Unlike conventional data analytic systems that employ time-consuming and expensive trial and error for identifying a resolution via elimination method, the present system 100 allows for generation of only a subset of test cases for accurately reproducing and resolving the chronic issue.
[0030] FIGs 2-3 depict a graphical representation of exemplary chronic issues that were automatically resolved using the system of FIG. 1. Particularly, FIG. 2 illustrates a graphical representation 200 of a random reboot issue reported in certain STBs. Since, there was no recognizable pattern, the issue remained unsolved using conventional STB troubleshooting for a long time. However, use of the progressive

analytics modeling by the system 100 identified the root cause as a rare exception condition that occurred when multiple tuners were in operation, the tuners were receiving a degraded signal, and hard disk occupancy was above a certain threshold.
[0031] Similarly, FIG. 3 illustrates a graphical representation 300 of a long¬standing ghost key press issue. However, use of the progressive analytics modeling by the system 100 identified the root cause as an abnormal working condition of the operation system when an amount of available memory for cache swap is lower than a particular threshold of about 15 megabytes (MB) and if the system had been running for more than 15.65 hours.
[0032] Further, FIG. 4 illustrates a schematic representation 400 of an exemplary process flow for providing automated chronic issue resolution in CPE using the system of FIG. 1. At step 1, one or more data tables 402 are provided to a feature engineering module 404. In one embodiment, the feature engineering module 404 evaluates each candidate parameter or feature based on how much variance it can explain, if the feature is truly independent, and/or if the features have low to negligible cross-correlation coefficient amongst themselves. If one or more of the features exhibit some degree of interdependency, the feature engineering module 404 combines the features into component space prior to the analysis.
[0033] Further, in step 2, an optimality selection module 406 prepares a final set of features after evaluating various metrics corresponding to the features analyzed by the feature engineering module 404. In one embodiment, the metrics considered by the optimality selection module 204 depend on the type of chronic problem being diagnosed. Additionally, the optimality selection module 406 uses deep pattern mining to identify the final set of features that are of greatest relevance to the occurrence of the chronic problem.

[0034] Further, at step 3, a training model 408 is executed with the final set of features identified at step 2. In certain embodiments, the training model 408 is selected based on the type of data being analyzed. By way of example, the training model 408 may include a dynamic decision tree, a neural network, and/or a Bayesian learning model. Subsequently, at step 4, a cost optimization module 410 is designed which determines suitable exit criteria for convergence of each training model identified at step 3. For example, when identifying occurrence of a random reboot issue depicted in FIG. 2, the cost optimization model may be designed based on a determined percentage of false positive data, a ratio of parameters involved to a total number of parameters that explains the chronic issue, and/or for maximizing a distance between clusters of the features that are indicative of the cause of chronic issue and clusters that are indicative of error-free operation.
[0035] However, in one embodiment, if the cost optimization module 410 determines two or more models to have similar accuracy, a new model is generated by combining a nearly equal performing model. Nearly equal performing models are those models whose performance parameters such as Type 1 error (False Positive) or Type 2 (False Negative) error remain within a specified threshold. If the new model performs better than the previous models, the new model is selected, else the least cost model from the nearly equal performing models is selected as the final model.
[0036] Further, at step 5, output parameters derived from the final model are converted into a suitable data format for easy interpretation. In case the model output contains component space parameters, component space to feature vector conversion is performed. Once the description and/or weightage of parameters are converted into an intuitive format, a TAS may easily derive inputs for the test cases needed for accurately reproducing the chronic issue.

[0037] By way of example, for the ghost press issue depicted in FIG. 3, one of the possible set of test cases which may be generated to recreate the chronic issue using the information from the model is as follows:
[0038] Run the box for 9 hours;
[0039] Perform a task which is kernel operation intensive;
[0040] Immediately run application which needs considerable about of run time memory; and
[0041] Observe the memory available window until memory is less than 15MB.
[0042] FIG. 5 illustrates an exemplary diagram 500 depicting a pattern of parameters identified using the system of FIG. 1 that cause one or more chronic issues in a connected device to aid in selection of suitable test cases for faster resolution of the issue. Each failure path 502 in the diagram 500 may be used as a potential input for selection of a suitable test case. In one embodiment, the diagram 500 may be used to select 4-8 test cases suitable for reproducing the issue repetitively and accurately. Reproduction of the chronic issue via such a test case validates the root cause analysis, thus allowing for identification of an appropriate resolution of the chronic issue.
[0043] In certain embodiments, embodiments of the system 100 and method 400 may further include a self-learning module that uses the various analyses and details of the chronic issue resolution as feedback for enhancing subsequent predictions by the system 100 and/or the method 400. An exemplary performance of the trained outcome model developed using the system and method described with reference to of FIGs. 1-4 in predicting chronic issues is depicted in FIG. 6.
[0044] Specifically, FIG. 6 illustrates a graphical representation 600 of statistics depicting a comparison of customer-reports of a chronic issue and instances of the

same chronic issue predicted using the trained outcome model developed using the system and method described with reference to of FIGs. 1-4. As shown in FIG. 6, the predictions of the occurrence of the chronic issue of "no image displayed" by trained outcome model largely match the day of the customer-reported occurrence of the issue. During the exemplary implementation, data for the issue reported on 13th June 2015 was not available in the database, thus no predicted issue on these dates. Moreover, absence of customer report of the occurrence of the issue on 3rd and 9th of June, for example, may be attributed to no report by customer.
[0045] Embodiments of the present systems and methods, thus, provide an efficient mechanism based on progressive analytical modeling of existing data to resolve problems, which are seemingly unreproducible in connected devices. Particularly, embodiments of the system and method described herein describe generation of a highly accurate problem prediction model using a collection of suitably selected deep mining, machine learning algorithms. When applied to the data sets collected in the context of the chronic issue, the model identifies a subset of parameters and their weightages directly responsible for causing the chronic issue. Subsequently, a test package may be generated based on the identified parameters and a corresponding representation for validating the identified root cause and aiding in subsequent issue resolution. Use of the progressive analytic modeling for automated chronic issue resolution reduces significant testing effort and expense, quickly resolves chronic issues, thus ensuring provision of superior services to the customers.
[0046] It may be noted that the foregoing examples, demonstrations, and process steps that may be performed by certain components of the present systems, for example, by the data integration framework 110, the analytics subsystem 112, the predictor modeling subsystem 120 may be implemented using hardware, firmware, and/or suitable code on a processor-based system, such as a general-purpose or a special-purpose computer. Moreover, even though FIG. 1 depicts the data integration

framework 110, the analytics subsystem 112, and the predictor modeling subsystem 120 as three different devices, in one embodiment, a single device may perform the functions of each of these subsystems. Alternatively, more than three devices may be used in lieu of the data integration framework 110, the analytics subsystem 112, and the predictor modeling subsystem 120. It may also be noted that different implementations of the present systems and methods may perform some or all of the steps described herein in different orders or substantially concurrently.
[0047] Additionally, various functions and/or method steps described in may be implemented in a variety of programming languages, including but not limited to Ruby, Hypertext Pre-processor (PHP), Perl, Delphi, Python, C, C++, or Java. Such code may be stored or adapted for storage on one or more tangible, machine-readable media, such as on data repository chips, local or remote hard disks, optical disks (that is, CDs or DVDs), solid-state drives, or other media, which may be accessed by the processor-based system to execute the stored code.
[0048] Although specific features of various embodiments of the present systems and methods may be shown in and/or described with respect to some drawings and not in others, this is for convenience only. It is to be understood that the described features, structures, and/or characteristics, and any subset thereof, may be combined and/or used interchangeably in any suitable manner in the various embodiments, for example, to construct additional assemblies and techniques for use various test automation systems.
[0049] While only certain features of the present systems and methods have been illustrated and described herein, many modifications and changes will occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

We claim:
1. A method for identifying a chronic issue, comprising:
processing data corresponding to one or more connected devices operable in a communications network to define one or more boundary conditions corresponding to the chronic issue associated with a connected device in the plurality of connected devices;
identifying one or more candidate parameters from the processed data that are determined to cause greatest variability of the data;
determining one or more issue predictor models that aggregate a suite of learning algorithms to evaluate effect of one or more of the identified candidate parameters on occurrence of the chronic issue;
iteratively developing an outcome model from one or more of the issue predictor models based on a statistical analysis of performance of each of the issue predictor models such that the outcome model satisfies one or more specified exit criteria so as to identify occurrence of a root cause of the chronic issue with a desired accuracy level; and
communicating the root cause of the chronic issue to an associated test automation system.
2. The method as claimed in claim 1, further comprising determining one or more test vectors, test scripts, test cases, or combinations thereof, based on the root cause of the chronic issue to accurately reproduce the chronic issue.
3. The method as claimed in claim 1, further comprising automatically selecting one or more test vectors, test scripts, test cases, or combinations thereof, from a repository based on the root cause of the chronic issue to accurately reproduce the chronic issue.

4. The method as claimed in claim 1, wherein the data comprises customer data, device data, operational parameters of the connected device, operational parameters of one or more of the plurality of connected devices, data logs related to occurrence of the chronic issue in the connected device and one or more of the plurality of connected devices, other issues associated with the connected device, and one or more remedial actions previously performed for the chronic issue.
5. The method as claimed in claim 1, wherein the boundary conditions are defined based on domain knowledge.
6. The method as claimed in claim 1, wherein identifying the candidate parameters comprises selecting the candidate parameters and their corresponding relevance based on their p-value, and wherein the one or more candidate parameters are identified in one or more of a feature space and component space.
7. The method as claimed in claim 1, wherein the suite of learning algorithms comprises a support vector machine, classification, pattern recognition, logistic regression, linear discriminant analysis, decision tree, neural network, and Bayesian algorithm.
8. The method as claimed in claim 1, wherein the specified exit criteria comprises maximizing a distance between one or more clusters of the candidate parameters that are indicative of the root cause of the chronic issue and one or more clusters of the candidate parameters that are indicative of error-free operation.
9. The method as claimed in claim 1, wherein iteratively developing the outcome model comprises selecting one of the issue predictor models as the outcome

model based on one or more of associated cost and accuracy, or combining two or more of the issue predictor models, which provide accuracy within a specified threshold.
10. The method as claimed in claim 1, further comprising communicating one or more of the data, the statistical analysis, and the root cause of the chronic issue as feedback to a self-learning module to improve prediction of a future occurrence of the chronic issue in one or more of the plurality of connected devices.
11. The method as claimed in claim 1, wherein identifying the candidate parameters comprises:
determining one or more of a number of levels, cro-correlation, and a probability distribution function for each of the candidate parameters for identifying connections between the candidate parameters that are determined to cause greatest variability of the data; and
combining the candidate parameters that show more than a define degree of cross-correlation into component space prior to the statistical analysis.
12. The method as claimed in claim 1, wherein iteratively developing the outcome model from one or more of the issue predictor models comprises defining an objective function for cost optimization corresponding to each of the issue predictor models.
13. The method as claimed in claim 1, wherein one or more of the candidate parameters, the issue predictor models, and the suite of learning algorithms is selected based on one or more of a type of chronic issue and a type of the identified candidate parameters.

14. A system for critical problem resolution, comprising:
a storage subsystem configured to store data corresponding to one or more connected devices operable in a communications network;
an analytics subsystem communicatively coupled to the storage subsystem and configured to:
process the data to define one or more boundary conditions corresponding to the chronic issue associated with a connected device in the plurality of connected devices;
identify one or more candidate parameters from the processed data that are determined to cause greatest variability of the data;
determine one or more issue predictor models that aggregate a suite of learning algorithms to evaluate effect of one or more of the identified candidate parameters on occurrence of the chronic issue;
iteratively develop an outcome model from one or more of the issue predictor models based on a statistical analysis of performance of each of the issue predictor models such that the outcome model satisfies one or more specified exit criteria so as to identify occurrence of a root cause of the chronic issue with a desired accuracy level; and
a test automation system communicatively coupled to one or more of the storage subsystem and the analytics subsystem, and configured to:
receive the identified root cause of the chronic issue from one or more of the storage subsystem and the analytics subsystem;
automatically select one or more test vectors, test scripts, test cases, or combinations thereof, from a repository based on

the root cause of the chronic issue to accurately reproduce the chronic issue.