M-Lean: An end-to-end development framework for predictive models in B2B scenarios

Abstract

The need for business intelligence has led to advances in machine learning in the business domain, especially with the rise of big data analytics. However, the resulting predictive systems often fail to maintain a satisfactory level of performance in production. Besides, for predictive systems used in business-to-business scenarios, user trust is subject to the model performance. Therefore, the processes of creating, evaluating, and deploying machine learning systems in the business domain need innovative solutions to solve the critical challenges of assuring the quality of the resulting systems. Applying machine learning in business-to-business situations imposes specific requirements. This paper aims at providing an integrated solution to businesses to help them transform their data into actions. The paper presents MLean, an end-to-end framework, that aims at guiding businesses in designing, developing, evaluating, and deploying business-to-business predictive systems. The framework employs the Lean Startup methodology and aims at maximizing the business value while eliminating wasteful development practices. To evaluate the proposed framework, with the help of our industrial partner, we applied the framework to a case study to build a predictive product. The case study resulted in a predictive system to predict the risks of software license cancellations. The system was iteratively developed and evaluated while adopting the management and end-user perspectives. It is concluded that, in industry, it is important to be aware of the businesses requirements before considering the application of machine learning. The framework accommodates business perspective from the beginning to produce a holistic product. From the results of the case study, we think that this framework can help businesses define the right opportunities for applying machine learning, developing solutions, evaluating the effectiveness of these solutions, and maintaining their performance in production.