Speaker "Tushar Sabharwal" Details Back



Machine learning | Big data | Artificial Intelligence Telecom Way


Machine Learning & Predictive Analysis Introduction Machine learning is often used to build predictive models by extracting patterns from large datasets. These models are used in predictive data analytics applications including price prediction, risk assessment, predicting customer behavior, and document classification. This introductory textbook offers a detailed and focused treatment of the most important machine learning approaches used in predictive data analytics, covering both theoretical concepts and practical applications. Technical and mathematical material is augmented with explanatory worked examples, and case studies illustrate the application of these models in the broader business context. Hypothesis 1 Mentioned approach presents the way to predict the best optimum configuration for deployment of Mediation software using machine learning and prediction algorithms. In the old conventional approach, optimum configuration needs to calculate manually on the basis of some inputs and static factor. We tried to find out the best optimum configuration using machine learning and prediction algorithems in the automated fashion and after some numbers of itrations results become more efficinet, we took a step forward and try to implement same solution using ML, prediction algorithems with the help of BIG DATA. After using the Big Data technologies, it very easy for us to manage data and data mining to required data. Problem Statement: In the current situaltion, Hardware requirment and bussiness logic needs to be developed manually, when such requirement and logic needs to be created, decision depend upon the some inputs and static factor. Initially when there is a little traffic, everything works fine so we need a system which will consider not only the rational points but the future considerable point as well. Proposed Solution: To achieve that we start gathering the data and implanted data mining techniques to it and parsed the data through a Machine learning logic and as a output a new list of required hardware and business logic come out from this. We did the lots of iterations to this and we got the more efficient business logics out of this as we get more and more data. After this, we do the data mining and data gathering using the Big Data technologies, result is still the same but the implantation and analysis time is reduced to two third of the last system. To achieve the same, we used machine learning, prediction algorithms and big data. Our next motive is to develop Intelligence around it using artificial intelligence, first mile stone is to develop a weak AI and then second mile stone is to develop a Strong AI around this system. Hypothesis 2 In the second use case, prediction based Hung state monitoring concept is created using machine learning, Big data and prediction algorithems, It also defines the new way of monitoring the system which is different from the conventional monitoring systems. In the present scenario, we use the High end computers which run the mission critical applications like web applications or any financial database which requires the 99.999% uptime means 4 min downtime in a year. We use the different High Availability solutions like Sun Clusters or Veritas Cluster or RHEL cluster but they also require downtime at the time of failover activity. In the present condition, we monitor the system statics and when any crucial parameter is go above the threshold value it triggers an alarm, Monitoring team send the same alarm to application or appropriate team and after analysis appropriate team take the action, this complete activity took around 5-10 min to complete and in between some time system got in to panic state and leads to mission critical application downtime. Proposed Solution: Proposed solution will recognize the system parameter usage pattern and intelligently predict the system behavior. On the basis of prediction algorithm we can take the most appropriate solution. Proposed solution is to design a client-server architecture based solution, which has the its own defined prediction algorithm, High level view of activities performed by this solution is as following, detailed description will be mentioned later After every N min, the same cycle performed 1) Monitor the system statistics like CPU, Memory and Load Average for the last M hours 2) Using the Predictive algorithm, validate if any of the system parameter will reach the threshold value 3) Using machine learning and patter matching, intelligent solution needs to be proposed 4) In case, Manual intervention was not performed, intelligently system defined action will be performed.


Test Specialist in Ericsson with 10 Years of experience Expertise Testing Product | Concept to Commercialization Responsible for designing, securing and monitoring deployment for telecom Products Involved in POCs for product transformation.