Customer Segmentation And Repeat Purchase Prediction Of Pharmacy Based On Data Mining

With the aging population and the impact of the COVID-19 pandemic,people are paying more attention to their physical health.Chain pharmacies have become the main place for people to buy medicines because of their rich variety of medicines and quality services.However,under the fierce market competition,pharmaceutical retail companies face increasingly more challenges.How to meet the different needs of consumers while ensuring the quality of drugs and services,and how to improve customer loyalty and repurchase rates have become important issues for the development of pharmaceutical retail companies.Therefore,establishing suitable customer segmentation and repurchase behavior prediction models has significant implications for companies to better understand customer needs and behavior,develop effective marketing strategies,and improve business performance.Customer segmentation is the process of dividing customers into different categories based on their similarities and differences,so that companies can develop targeted marketing strategies for different types of customers.Customer repurchase behavior prediction,on the other hand,aims to forecast customers’ future purchasing tendencies,which can help companies develop targeted promotional strategies to improve customer repurchase rates and loyalty.Many researchers at home and abroad have begun to focus on customer segmentation and repurchase behavior prediction,and have achieved some research results.Data mining algorithms are effective tools that can extract valuable information and patterns from large amounts of data.Clustering algorithms are unsupervised learning algorithms in data mining,which can group samples in a data set according to a certain similarity or distance metric.Ensemble learning is an effective machine learning method that combines the results of multiple base classifiers to improve classification accuracy and robustness.The research results of the customer segmentation model showed that,compared with the K-means algorithm,the Mini-Batch K-means algorithm significantly reduced training time while achieving good clustering results,with a speed increase of nearly ten times,making it suitable for processing large datasets.For data with different eccentricities,using the initialization clustering center method of init= "Huang" can yield better results when applying the K-prototype clustering method.From the CH coefficient and DBI value of K-means and K-modes,the categorical and numerical variables of the customer data set are somewhat correlated,but the correlation is not strong,indicating that customer purchasing behavior is influenced by multiple factors for this data set.Regarding the customer repurchase prediction model,the research results showed that the ensemble learning model had good classification performance in predicting customer purchasing behavior,with the Ada Boost model having superior accuracy,precision,and recall rates compared to the random forest and gradient boosting decision tree models.From the feature importance ranking output of the ensemble learning model,the time since the last purchase,the number of purchases in the last six months,the average purchase days,the frequently purchased product categories,and age were found to have a significant impact on whether customers repurchase.Enterprises should pay attention to changes in customer behavior in these areas,combine with customer segmentation results,and adopt appropriate marketing and maintenance strategies,such as providing personalized services,coupons,or promotional activities,to enhance customer loyalty,improve customer satisfaction and enterprise revenue.