Artificial intelligence, machine learning, and deep learning have achieved great success in the fields of computer vision and natural language processing, and then extended to various fields, such as biology, chemistry, and civil engineering, including big data in the field of logistics. Therefore, many logistics companies move towards the integration of intelligent transportation systems. Only virtual and physical development can support the sustainable development of the logistics industry. This study aims to: 1.) collect timely information from the block chain, 2.) use deep learning to build a customer database so that sales staff in physical stores can grasp customer preferences, and 3.) integrate Generative Adversarial Network analysis and logistics truck delivery route analysis. This study will introduce new logistics technology development and innovative smart service structure, covering front-end Internet of Things sensing, mobile application apps, and back-end massive data analysis platform to promote the self/intelligence of logistics. Artificial intelligence for customer preference analysis is used, and images are automatically distributed through the system to reduce labor costs and increase sales. The proposed method is feasible, and it also achieves the push system of information transmission in transportation. Thus, logistics transportation cost transmission is reduced, thereby intelligently pushing self-promotion in marketing activities.
Nowadays, the development of e-commerce has moved towards Online to Offline (O2O) integration. It has become a mainstream for consumers to conduct mobile shopping through mobile phones or computers. This approach has made shopping a quick and convenient experience; thus, e-commerce platforms combine with physical stores to adopt a multichannel approach for product promotion. O2O integration can achieve the goal of omni-channel marketing. While consumers can make purchases at any time and place in the virtual market, the physical stores can provide consumers with services, such as logistics, pick-up, and product experience. This approach can fully combine the advantages of O2O to maximize profits and diversify customer base. The main customer base of traditional marketing includes the peripheral business districts and shop consumption customers. The influence is limited to the customers in the business district. With O2O integration, the source of customers can be diversified, and the physical stores can combine multiple e-commerce players to allow more diversified product distribution channels. When an online brand establishes a physical channel, three disadvantages are found compared with the virtual market, as follows: 1.) costs of rent, 2.) surplus inventory, and 3.) product profit. The major problem is the lack of control on the sales of goods, leading to the stockpiles of slow sale products occupying the showroom space and causing invalid inventory. In the virtual market, goods can only be viewed from photos or video clips; consumers cannot actually experience or learn more about the content of the goods.
In view of this, virtual and physical integration can incorporate the convenience of virtual market and the actual experience of physical stores. To construct a complete virtual and physical integration project and allow the industry to set up physical stores based on their business strategies, BlockChain technologies can be employed to link the data from various platforms, such as: sales data, inventory status, etc. In addition to online product recommendations, such as recommendations on social media, or e-mails, etc., physical stores can analyze customer needs through the recommendation system to strengthen customer relationships. Thus, the system needs to accurately analyze customer consumption needs to improve sales performance. However, the current recommendation system can only recommend products online for a single information platform. Physical stores thus cannot connect and aggregate the membership data on various information platforms; the system thus cannot accurately analyze consumer behavior. The precision marketing through the analysis of consumer behavior across e-commerce operators is the only path to Retail 4.0 era. Therefore, to establish capacities in consistent experience, the firms need to integrate data from various platforms and observe customer consumption behaviors through third-party channels to achieve data availability.
Brent [
The construction of a customer marketing system was proposed. Each customer has different product preferences; thus, the system generates a set of mosaic based on the customer’s preference and sends it to the customer. If the sales staff can grasp the customers’ consumption behavior and predict their purchase time, then the overall sales volume can be increased. A customer marketing system using blockchain information was developed. It is used to analyze customer consumption patterns through deep learning algorithms and predict customer product shopping cycles. The proposed smart contract also combines the pickup models of a physical store and automatically notifies the sales staff and customers. A customer marketing system based on deep learning was constructed using long short-term memory (LSTM) to analyze consumer consumption patterns and GAN to generate picture. The system needs to collect information from the blockchain from time to time for analysis. In addition to analyzing the customers’ own consumption behavior, the customer marketing system in the study can judge the customer’s product preferences on the basis of the similarity with other customers when the initial customer data are small. The forecast time cycle of purchasing and the analysis of future product trends were also established. A marketing system is developed, and the data of O2O were integrated to allow a more diversified business marketing strategy and customer base to increase the overall sales volume. This study uses a customer marketing system to construct different customers’ product preferences. Using this system can reduce the time for customers to choose products and allow manufacturers to accurately target customer needs. This study uses deep learning to judge customer needs from the blockchain and establishes a marketing system that is very helpful to both consumers and manufacturers.
Dongyuan et al. [
Qiang et al. [
Qiang et al. [
This study mainly employed BlockChain to integrate the virtual and physical data and adopted LSTM to carry out the customer purchase cycle; social network software was then used to push the marketing advertisement images generated by GAN. The method introduced in the study facilitated automatic marketing to increase the firm’s sales volume.
This section introduces the system architecture of this study and Bilinear Pairings.
The system architecture of this study is shown in
The study employed Bilinear Pairings technology to construct the network security mechanism of BlockChain, which enhance the fast search and security of BlockChain. Suppose Bilinear: Non-degeneracy: Computable: There exists an efficient algorithm to compute
In Scott [
This section described 4.1 BlockChain search; 4.2 LSTM product prediction; and 4.3 GAN image generation.
First, each store use IBC to encrypt customer transaction data and transmit customer-related data to the BlockChain. The data are
After collecting consumer data, data analysis is required. The system mainly analysis the following information of consumers: 1. Consumption amount; 2. Frequency of consumption; 3. Methods of payment; 4. Categories of product purchased. The system will analyze the consumption amount to understand the average value and change of the purchase amount each time, which is calculated as follows:
(1) The consumer’s
(2) Next, calculate the change of consumption amount; the system employs the variance to calculate. The calculation is
Next, the number of data is used as consumption frequency (
This study employed Long Short-Term Memory (LSTM) to conduct predict the sales of products. The calculation is as follows: The study collects the daily sales volume of each product as First, calculate the input part Next, calculate the forget gate as Next, calculate the output layer as
Next, the system will calculate the expected sales volume of each product the next day. On one hand, one can understand the product sales status for review or related promotion activities.
The study predicts the purchase cycle of customers and automatically notify high frequency customers the related promotion on the frequently purchased products to encourage customers to visit the physical stores and make further purchase and experiment. In addition, this study constructs future sales forecasts of products, using LSTM for prediction. The study takes product purchase time to sequence the daily sales volume to forecast the future product sales; so that the firms can have a better handle on the product sales and inventory to facilitate the dispatch.
The study first establishes a fixed term database CSQL. CSQL will create comments corresponding to the predicted value of the output layer, such as: high consumption regular customers, credit card payment, purchasing skin care products, etc.; then, the analysis is as follows: The first-order eigenvalue calculation model is Suppose the predicted value is calculated as The system inputs
4. Next, hyperbolic tangent, Rectified Linear Unit, and excitation function are employed. The gradient of Rectified Linear Unit is a constant 5. The predicted value after the input of the input layer is 6. The system will correspond to the word database CSQL according to the predicted value
The establishment of a customer marketing system in this study helps marketers to have better understanding of consumer preferences. Although many POS systems today can list customer transaction details, it is difficult for marketers to grasp customers from transaction details in a short time. Therefore, this study proposes a customer marketing system which adopts deep learning to extract feature values and displays them on the Web through predictive values that correspond to relative words, which helps marketers better understand, from the results, customer spending habits and preferences. The system integrates the transaction details of the virtual market and can better grasp the customer needs to achieve the goal of virtual and physical (O2O) integration. To provide customers with more special offers, the system uses GAN for image synthesis technology. The system presets the background image as
The study performed LSTM prediction using a total of 1,000 customer data. As shown in
Nowadays customer shopping has moved from traditional stores to online shopping. Many customers make online purchase through mobile devices. Under the competition of e-commerce, if you can grasp the customer’s consumption interest and cycle, you can get the customer’s willingness to buy your products. This study proposed a marketing customer system which mainly uses LSTM for cycle prediction to understand a customer’s purchasing cycle; then uses GAN to perform marketing product synthesis and pushes the message through social media or SMS. According to the experiment results, the LSTM prediction accuracy is about 98%. From the results, it can be learned that the method proposed in the study is feasible; the method proposed in the study can help stores conduct smart marketing and improve their product sales performance.
The authors would like to thank Intelligent Automation & Soft Computing for their help in preparing this manuscript for publication.