Project Name : Capacity ceiling forecasting - Capacity and Performance Management

Technology : Python/Tableau/SQL/Excel

Client : Barclays (Investment Banking)

Project Description:

Developed regression model from scratch for analysing performance of servers and different investment banking applications. The model can forecast the capacity ceiling (number of transactions application can processed without breaching service level agreement) of the servers.

Responsibilities:

• Data visualization using Tableau the data and perform initial analysis on available data.
• Data pre-processing using encoding methods like probability ratio and log of weight ratio.
• Developed baseline training model using linear regression, random forest regression and support vector regression to forecast capacity ceiling of application.
• Evaluate developed baseline models on the basis of evaluation metrics.
• Hyper parameter tuning of baseline model using evaluation metrics and selection of best model and persist for future reference.
• Predicted the capacity ceiling value using final model, saved to database, display on our CPM dashboard, generating capacity, performance graphs using Python and displayed on dashboard.
• Develop and publish reports based on client requirements using data visualization with Tableau i.e. volume trend, volume forecast report.

Project Name : Exception detection for Memory, CPU, IO - Capacity and Performance Management

Technology : Python/Tableau/SQL/Excel

Client : Barclays (Investment Banking)

Project Description:

Developed anomaly detection algorithm for identification of exceptions based on past data in the performance as scope of the Capacity and Performance management has increased to Barclays international. Currently working on Deep learning mechanism for anomaly detection system.

Responsibilities:

• Developed baseline classification training models using logistic regression, decision tree mechanism, Random forest and support vector machines to identify exceptions.
• Evaluate developed baseline models on the basis of evaluation metrics applied boosting mechanism on random forest algorithm, and also evaluate best parameters using evaluation metrics.
• Hyper parameter tuning using best parameters evaluation metrics and selection of best model and persist for future reference.