本文選題：消費者意圖 + 瀏覽��； 參考：《五邑大學》2016年碩士論文

【摘要】：隨著互聯(lián)網(wǎng)應(yīng)用在商業(yè)領(lǐng)域的快速普及,用戶的需求體驗成為互聯(lián)網(wǎng)發(fā)展的驅(qū)動力。以電子商務(wù)系統(tǒng)為代表的社交網(wǎng)絡(luò)不斷發(fā)生新的變化。電子商務(wù)系統(tǒng)在為用戶提供越來越多選擇的同時,其結(jié)構(gòu)也變得更加復(fù)雜,用戶經(jīng)常會迷失在大量的商品信息空間中,無法順利找到自己需要的商品。電子商務(wù)推薦系統(tǒng)直接與用戶交互,模擬商店銷售人員向用戶提供商品推薦服務(wù),幫助用戶找到所需商品,從而順利完成購買過程。隨著電子商務(wù)系統(tǒng)規(guī)模的進一步擴大,電子商務(wù)推薦系統(tǒng)也在飛速的發(fā)展。目前主流的電子商務(wù)推薦系統(tǒng)有:基于協(xié)同過濾算法的電子商務(wù)推薦系統(tǒng);基于用戶統(tǒng)計算法的電子商務(wù)推薦系統(tǒng);基于知識發(fā)現(xiàn)算法電子商務(wù)推薦系統(tǒng)和基于效用的電子商務(wù)推薦系統(tǒng)。但這些推薦系統(tǒng)存在著各自的局限性,因此在電子商務(wù)推薦系統(tǒng)中,消費者意圖的識別承擔著越來越重要的作用。消費者意圖的識別對于電子商務(wù)商品推薦、熱點引流商品選取、網(wǎng)站布局以及鏈接的設(shè)置有至關(guān)重要的影響。目前的大部分研究都認為意圖是靜態(tài)的,即特定的意圖是伴隨著特定的環(huán)境的,因而在特定的環(huán)境下有特定的變化。然而,消費者在電子商務(wù)活動中訪問和選購商品時的不確定性告訴我們消費者的意圖可表現(xiàn)為多種形態(tài),并且是多階段發(fā)展的。因此,本研究用蟻群算法中的螞蟻來表示消費者,通過螞蟻對信息素的趨好性來模擬消費者的瀏覽、收藏、加入購物車和購買行為的意圖。因為消費者意圖為表現(xiàn)為商品的客觀屬性與消費者的主觀感受的匹配,所以,我們將信息素表示為商品的客觀屬性和消費者感知能力的內(nèi)積,其值即為表示消費者意圖的信息素的濃度。我們把這種信息素稱為感知信息素,用來表示消費者意圖。這樣就可以通過蟻群算法來呈現(xiàn)消費者意圖的動態(tài)性和不確定性。然后,本研究通過NetLogo仿真實驗以獲取數(shù)據(jù),再以神經(jīng)網(wǎng)絡(luò)來識別和驗證消費者的瀏覽、收藏、加入購物車和購買意圖。實驗結(jié)果表明:在95%的顯著性水平下,本研究所提出的模型將意圖預(yù)測的準確度從48%提升到67%左右,可以更加準確的向用戶推薦商品,具有良好的現(xiàn)實意義。
[Abstract]:With the rapid popularity of Internet applications in the business field, user needs experience has become the driving force of the development of the Internet. Social networks, represented by e-commerce systems, are constantly undergoing new changes. Electronic commerce system provides more and more choices for users at the same time, its structure becomes more complex, users often lose in a large number of commodity information space, can not find their own needs of goods. The E-commerce recommendation system directly interacts with the user, simulates the shop salesperson to provide the product recommendation service to the user, helps the user to find the needed product, thus completes the purchase process smoothly. With the further expansion of e-commerce system, e-commerce recommendation system is also developing rapidly. At present, the main E-commerce recommendation system includes: E-commerce recommendation system based on collaborative filtering algorithm, e-commerce recommendation system based on user statistics algorithm, and electronic commerce recommendation system based on user statistics algorithm. E-commerce recommendation system based on knowledge discovery algorithm and utility-based e-commerce recommendation system. However, these recommendation systems have their own limitations, so the recognition of consumer intention plays an increasingly important role in e-commerce recommendation systems. The identification of consumer intention is of great importance to the recommendation of e-commerce products, the selection of hot spots, the layout of websites and the setting of links. Most of the current studies suggest that the intention is static, that is, the specific intention is accompanied by the specific environment, so there is a specific change in the specific environment. However, the uncertainty when consumers visit and choose goods in e-commerce activities tells us that the intention of consumers can be expressed in a variety of forms, and is multi-stage development. Therefore, the ants in ant colony algorithm are used to represent consumers, and the intention of consumers to browse, collect, add shopping cart and purchase behavior is simulated by ants' readability to pheromone. Because consumers intend to match the objective attributes of commodities with the subjective feelings of consumers, we express pheromones as the inner product of the objective attributes of commodities and the perception ability of consumers. Its value is the concentration of pheromone that represents the intention of the consumer. We call this pheromone a perceptual pheromone, which is used to express consumer intention. In this way, ant colony algorithm can be used to show the dynamic and uncertainty of consumer intention. Then, through the NetLogo simulation experiment to obtain the data, and then the neural network to identify and verify the browsing, collecting, adding shopping cart and purchase intention. The experimental results show that the proposed model can improve the accuracy of intention prediction from 48% to 67% under the significance level of 95%. It can recommend products to users more accurately and has good practical significance.
【學位授予單位】：五邑大學
【學位級別】：碩士
【學位授予年份】：2016
【分類號】：F713.55;F713.36;TP18
，

本文編號：2075354