Predictive Analytics and Consumer Loyalty
Related works
Various efforts have been made to build an effective prediction model for retaining customers using different techniques. To better understand how Many studies have built their own predictive models suggested by Oladapo et al. Logistic regression model design, a good model of customer data to predict customer retention in a telecommunications company with 95.5% accuracy. This model predicts customer retention based on billing, value-added services, and SMS service issues.
Aluri et al. have focused on using machine learning to determine the value of customers in the hospitality sectors of customers, such as restaurants and hotels, by engaging dynamic customers with the loyalty program brand. Their results also show that automated learning processes excel in identifying customers with greater value in specific promotions. They have deepened the practical and theoretical understanding of automated learning in the value chain of customer loyalty, in a structure that uses a dynamic model for customer engagement.
Wiaya and Gersang predict customer loyalty at the National Multimedia Company of Indonesia, using three data mining algorithms, to form a customer loyalty classification model, namely: C4.5, Naive Bayes and Nearest Neighbor. These algorithms were applied to the set of data contained 2269 records and 9 attributes to be used. By comparing the analysis models, the C4.5 algorithm with its own data set segment (80% for training data and 20% of test data) has the highest accuracy results of 81.02% compared to algorithms and other data segments. In the attribute analysis, the disconnection attributes (the attribute that is interpreted as the reason why customers have stopped) get the most influential attribute on the accuracy of the results in the data extraction process to predict customer loyalty. This article does not discuss the algorithms of features selection, methods of obtaining important features, and its impact on model accuracy.
Wong and Wei presented a research to develop a tool to analyze customer behavior and predict their upcoming purchases from Air Travel Company. They provided an integration tool between data mining Pricing for competitors, customer segmentation and predictive analysis. Results In customer segmentation analysis, 110,840 clients are identified and segmented based on their purchasing behavior. Customers' profiles are split using a weighted RFM model, and customer purchasing behavior is analyzed in response to competitor price changes. The following destinations are expected for high-value customers identified using pre-link rules and custom packages promoted to targeted customer segments.
Moedjionom et al. have predicted customer loyalty in a multimedia services company, offering many services to win the market. This research contribution is to use data related to the segmentation and splitting of potential customers based on the RFM model, then applying the classification, Proportion of accuracy in customer loyalty rating research. Although the C4.5 algorithm with the k-mean segmentation give a better result, there are some important action that can be added to the search: using optimization algorithm to select the features or to adjust the value of the label to obtain a more accurate model.
Kaya et al. have built a predictive model based on spatial, temporal and optional behavioral features using individual transaction logs. Our results show that proposed dynamic behavioral models can predict change decisions much better than demographics-based features and that this effect remains constant across multiple data sets and different definitions of customer leakage. They have examined the relative importance of different behavioral features in predicting leakage, and how predictive power differed across different population groups.
Cheng and Sun have viewed other application of the RFM model (named TFM) to identify high-value customers in the communications industry. Use three main features to describe users who have accumulated a greater amount of service time (T), often purchase 3G services (F) and create large amounts of invoices per month (M).
This study proposes a comprehensive CRM strategy framework that includes customer segmentation and behavior analysis, using a dataset that contains about 500 million (full dataset in syriatel company). Al Janabi and Razaq used intelligent big data analysis to design smart predictors for customer churn in the telecommunication industry. The goal of this research maintain customers and improve the level of revenue. The proposed system consists of three basic pashas: First Phase: an understanding of the company's data. This phase focuses on the initial processing of data that is fragmented and unbalanced. They addressed the problem of imbalance by building the DSMOTE algorithm. Second Phase: construct a GBM-based predictor after it was developed, replace its decision-making part, which is (DT) with a (GA) algorithm. The impact of this is able to overcome DT problems and reduce time implementation. Third Stage: The accuracy of the predictor results was verified by using the matrix of the conflict matrix. A comparison was made between the traditional method of initial treatment, which is SMOTE, DSMOTE in terms of error rate and accuracy. GBM-GA method has higher Accuracy than GBM.
One of the biggest challenges of the current big data landscape is our inability to process vast amounts of information in a reasonable time. Reyes-Ortiz et al. explored and compared two distributed computing frameworks implemented on commodity cluster architectures: MPI/OpenMP on Beowulf that is high-performance oriented and exploits multi-machine/multi- core infrastructures, and Apache Spark on Hadoop which targets iterative algorithms through in-memory computing. The Google Cloud Platform service was used to create virtual machine clusters, run the frameworks, and evaluate two supervised machine learning algorithms: KNN and Pegasos SVM. Results obtained from experiments with a particle physics data set show MPI/OpenMP outperforms Spark by more than one order of magnitude in terms of processing speed and provides more consistent performance. However, Spark shows better data management infrastructure and the possibility of dealing with other aspects such as node failure and data replication.
There are several studies in the field of communication that deal with predicting the age and gender of the customer in big data platform by analyzing their personal data, including the study presented by Zaubi. Where he designed a model using a reliable data set of 18,000 users provided by SyriaTel Telecom Company, for training and testing. The model applied by using big data technology and achieved 85.6% accuracy in terms of user gender prediction and 65.5% of user age prediction. The main contribution of this work is the improvement in the accuracy in terms of user gender prediction and user age prediction based on mobile phone data and end-to-end solution that approaches customer data from multiple aspects in the telecom domain.
Other studies have also dealt with the prediction of customer churn in telecom using machine learning in big data platform, including the study presented by Ahmad. The main contribution of his work is to develop a churn prediction model which assists telecom operators to predict customers who are most likely subject to churn. The model developed in this work uses machine learning techniques on big data platforms and builds a new way of features' engineering and selection. In order to measure the performance of the model, the Area Under Curve (AUC) standard measure is adopted, and the AUC value obtained is 93.3%. Another main contribution is to use customer social networks in the prediction model by extracting Social Network Analysis (SNA) features. The use of SNA enhanced the performance of the model from 84 to 93.3% against the AUC standard.
With regard to how some studies approached customer value analysis, retention, and loyalty. A study did not apply to big data as it studied all customers according to some features and using a method of machine learning (a logistic regression model) to show the role of machine education in retaining and increasing customer loyalty. Machine learning was implemented in a major hospitality location and compared to traditional methods to determine customer value in the loyalty program. In the study predict customer loyalty at the National Multimedia Company of Indonesia, using three data mining algorithms, These algorithms were applied to the set of data obtained are 2269 records and contain 9 attributes to be used. By comparing the analysis models, the C4.5 algorithm with its own data set segment has the highest accuracy results of 81.02% compared to algorithms and other data segments. In my study, a model is built to increase customer loyalty predictions based on the new TFM methodology and machine learning. My experiences were demonstrated that TFM most appropriate for the telecom sector than RFM. The concept of the TFM is adjusted, where T is the sum of the duration of calls and the periods of internet sessions during a certain period. The set of data obtained is 127 million records and contains 220 features to be used. Binary and multi-classification are applied. After comparing the classifiers, the Gradient-boosted-tree classifier was found to be the best in binary and Random Forest Classifier is the best in multi-classification.