Mohammad Javad Jafari; M. J. Tarokh; Paria Soleimani
Abstract
Customer churn prediction has been gaining significant attention due to the increasing competition among mobile service providers. Machine learning algorithms are commonly used to predict churn; however, their performance can still be improved due to the complexity of customer data structure. Additionally, ...
Read More
Customer churn prediction has been gaining significant attention due to the increasing competition among mobile service providers. Machine learning algorithms are commonly used to predict churn; however, their performance can still be improved due to the complexity of customer data structure. Additionally, the lack of interpretability in their results leads to a lack of trust among managers. In this study, a step-by-step framework consisting of three layers is proposed to predict customer churn with high interpretability. The first layer utilizes data preprocessing techniques, the second layer proposes a novel classification model based on supervised and unsupervised algorithms, and the third layer uses evaluation criteria to improve interpretability. The proposed model outperforms existing models in both predictive and descriptive scores. The novelties of this paper lie in proposing a hybrid machine learning model for customer churn prediction and evaluating its interpretability using extracted indicators. Results demonstrate the superiority of clustered dataset versions of models over non-clustered versions, with KNN achieving a recall score of almost 99% for the first layer and the cluster decision tree achieving a 96% recall score for the second layer. Additionally, parameter sensitivity and stability are found to be effective interpretability evaluation metrics.
Zahra Jiryaei Sharahi; Yahia Zare Mehrjerdi; Mohammad Saleh Owlia; Masoud Abessi
Abstract
In a data-driven decision-making process, there are various types of data that should be thoroughly processed and analyzed. Data mining is a well-recognized method to obtain such information by analyzing data and transforming it into actionable insights for further use. Among the various data mining ...
Read More
In a data-driven decision-making process, there are various types of data that should be thoroughly processed and analyzed. Data mining is a well-recognized method to obtain such information by analyzing data and transforming it into actionable insights for further use. Among the various data mining techniques such as classification, clustering, and association rules, this research focused on classification techniques and presented an innovative regression-based learning approach in the decision tree (DT) models. DT algorithms are easy-to-understood and can work with different data types including continuous, discrete, and non-numerical. Despite a large number of existing studies, which attempt to enhance the performance of the DT models, there is still a gap in accurately extracting knowledge from databases. In this research, this issue is addressed by exploiting regression and coefficient of determination (R2) methods in a DT. The proposed tree provides new insights in the following aspects: split criterion, handling continuous and discrete variables, labeling leaf node, pruning process by stopping criteria and tree evaluation. The superiority of the proposed algorithm is demonstrated using a real-world hospital database and a comparison with existing approaches is provided. The results showed that the proposed algorithm outperforms the existing methods in terms of higher accuracy and lower complexity.
Mohammad Alipour-Vaezi; Reza Tavakkoli-Moghadaam; Mina Samieinasab
Abstract
Since human societies have endured massive financial disruptions and life losses after the outbreak of the COVID-19 pandemic, it is critical to eliminate this disease as soon as possible. Today, the invention of the COVID-19 vaccine made this objective more reachable. But unfortunately, the suppliant ...
Read More
Since human societies have endured massive financial disruptions and life losses after the outbreak of the COVID-19 pandemic, it is critical to eliminate this disease as soon as possible. Today, the invention of the COVID-19 vaccine made this objective more reachable. But unfortunately, the suppliant of the vaccines is limited. Hence, to prevent further lethal harms, it seems rational to use a scientific method for vaccine allocation. This study proposes a method for prioritizing the patients based on their level of life-threatening danger according to the proven risk factors (e.g., age, sex, pregnancy, and underlying diseases) of the COVID-19. That is a new data-driven decision-making method for patients’ classification based on their health condition information using several machine learning algorithms. In this method, vaccine applicants are classified into four classes. The scheduling of vaccine distribution would be conducted based on the results of this classification. Furthermore, a real-life case study is also investigated through the proposed method for better illumination in this paper. The vaccine distribution schedule of the real-case study has been performed with 94% accuracy. It should be mentioned that the main achievement of this research is to design a new efficient method for a vaccine distribution schedule.
M. J. Tarokh; Mahsa EsmaeiliGookeh
Abstract
The present study attempts to establish a new framework to speculate customer lifetime value by a stochastic approach. In this research the customer lifetime value is considered as combination of customer’s present and future value. At first step of our desired model, it is essential to define ...
Read More
The present study attempts to establish a new framework to speculate customer lifetime value by a stochastic approach. In this research the customer lifetime value is considered as combination of customer’s present and future value. At first step of our desired model, it is essential to define customer groups based on their behavior similarities, and in second step a mechanism to count current value, and at the end estimate the future value of customers. Having a structure in modeling customer churn is also important to have complete customer lifetime value computation. Clustering as one of data mining techniques is practiced to help us analyze the different groups of customers, and extract mathematical model to count the customers value. Thereafter by using Markov chain model as stochastic approach, we predict future behavior of the customer and as a result, estimate future value of different customers. The proposed model is demonstrated by the customer demographic data and historical transaction data in a composite manufacturing company in Iran.