Demo
Recently, businesses produce large volumes at varying dimensions of data due to the effects of rapid digitalization and multi-channel marketing practices. The data that is produced and stored in relational databases creates a portfolio that is fit to produce a lot of value for businesses with regards to customer relationship management software works.
To benefit from this big data, machine learning projects are being developed. In machine learning projects, using all the data related to the goal is a common tendency. Yet, as a result of this tendency, the training dataset growing too large is not sufficient for machine learning algorithms to create a well enough performing machine learning model.
One of the solutions for this issue in Computer Science is feature extraction under the subject of dimensionality reduction. Within this project, campaign success prediction accuracy was worked on to be improved regarding the customer relationship management software named Next4biz CRM of Next4biz software company.
Within this project, a working data was created as a result of merging the campaign notification records sent to customers, which is stored on the Next4biz CRM database, and other related meaningful data stored. Campaign filters, which are easily created thanks to the user-friendly design of Next4biz CRM, were also included in this project.
Since each of these filters was used as a feature, over six thousand meaningful features were created. With these features, a classification study was conducted to determine whether a campaign would be successful on an individual customer basis.
The Principal Component Analysis (PCA) approach was used, which is one of the feature extraction methods aiming to increase prediction accuracy. When results were inspected, a satisfying amount of increase in accuracy was found.
