A Linguistic Intuitionistic Cloud Decision Support Model with Sentiment Analysis for Product Selection in E-commerce

Abstract

Online product reviews significantly impact the online purchase decisions of consumers. However, extant decision support models have neglected the randomness and fuzziness of online reviews and the interrelationships among product features. This study presents an integrated decision support model that can help customers discover desirable products online. This proposed model encompasses three modules: information acquisition, information transformation, and integration model. We use the information acquisition module to gather linguistic intuitionistic fuzzy information in each review through sentiment analysis. We also apply the information transformation module to convert the linguistic intuitionistic fuzzy information into linguistic intuitionistic normal clouds (LINCs). The integration module is employed to obtain the overall LINCs for each product. A ranked list of alternative products is determined. A case study on Taobao.com is then provided to illustrate the effectiveness and feasibility of the proposal, along with sensitivity and comparison analyses, to verify its stability and superiority. Finally, conclusions and future research directions are suggested.

Appendix A: Formula of LINCAA and LINCNWBM Operators

Let $\gamma_{ij}^{k} = \left( {\left( {Ex_{ij}^{k} ,En_{ij}^{k} ,He_{ij}^{k} } \right),\left( {ex_{ij}^{k} ,en_{ij}^{k} ,he_{ij}^{k} } \right)} \right)$$\left( {k = 1,2, \cdots ,q_{i} } \right)$ be a collection of transformed LINCs from the kth online review concerning alternative A_i under product feature C_j. Let $U = [ {\text{X}}_{\hbox{min} } , {\text{X}}_{\hbox{min} } ]$ be an effective domain. Note that the weight of each online review is equal to ${1 \mathord{\left/ {\vphantom {1 {q_{i} }}} \right. \kern-0pt} {q_{i} }}$. The overall evaluation value $\gamma_{ij}^{{}} = \left( {\left( {Ex_{ij}^{{}} ,En_{ij}^{{}} ,He_{ij}^{{}} } \right),\left( {ex_{ij}^{{}} ,en_{ij}^{{}} ,he_{ij}^{{}} } \right)} \right)$ concerning alternative A_i under product feature C_j can be constructed by aggregating all $\gamma_{ij}^{k}$ of each online review with the LINCAA operator.

$$ \begin{aligned} & LINCAA\left( {\gamma_{ij}^{1} ,\gamma_{ij}^{2} , \ldots ,\gamma_{ij}^{{q_{i} }} } \right) = \mathop \oplus \limits_{k = 1}^{{q_{i} }} \frac{1}{{q_{i} }}\gamma_{ij}^{k} \\ & \quad = \left( {X_{\hbox{max} } \left( {1 - \prod\limits_{k = 1}^{{q_{i} }} {\left( {1 - \frac{{Ex_{ij}^{k} }}{{X_{\hbox{max} } }}} \right)}^{{\frac{1}{{q_{i} }}}} } \right),\;\left( {\sum\limits_{k = 1}^{{q_{i} }} {\frac{1}{{q_{i} }}\left( {En_{ij}^{k} } \right)^{2} } } \right)^{{\frac{1}{2}}} ,\;\left( {\sum\limits_{k = 1}^{{q_{i} }} {\frac{1}{{q_{i} }}\left( {He_{ij}^{k} } \right)^{2} } } \right)^{{\frac{1}{2}}} ,} \right. \\ & \quad \left. {X_{\hbox{max} } \left( {\prod\limits_{k = 1}^{{q_{i} }} {\left( {\frac{{ex_{ij}^{k} }}{{X_{\hbox{max} } }}} \right)}^{{\frac{1}{{q_{i} }}}} } \right),\;\left( {\sum\limits_{k = 1}^{{q_{i} }} {\frac{1}{{q_{i} }}\left( {en_{ij}^{k} } \right)^{2} } } \right)^{{\frac{1}{2}}} ,\;\left( {\sum\limits_{k = 1}^{{q_{i} }} {\frac{1}{{q_{i} }}\left( {he_{ij}^{k} } \right)^{2} } } \right)^{{\frac{1}{2}}} } \right); \\ \end{aligned} $$

(10)

The overall LINC information of each alternative product $\gamma_{i}^{{}}$ can be calculated by aggregating all $\gamma_{ij}^{{}}$ with the LINCNWBM operator by considering the interrelationships among product features. The weight vectors of product features are obtained by using the entropy-weighted method, denoted as $w = \left( {w_{1} ,w_{2} , \ldots ,w_{m} } \right)^{\text{T}}$.

$$ \begin{aligned} & LINCNWBM_{w}^{p,q} \left( {\gamma_{i1} ,\gamma_{i2} , \ldots ,\gamma_{im} } \right) = \left( {\mathop \oplus \limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} \frac{{w_{i} w_{j} }}{{1 - w_{i} }}\left( {\gamma_{ij}^{p} \otimes \gamma_{il}^{q} } \right)} \right)^{{\frac{1}{p + q}}} \\ & \quad = \left( {X_{\hbox{max} } \left( {1 - \prod\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\left( {1 - \left( {\frac{{Ex_{ij} }}{{X_{\hbox{max} } }}} \right)^{p} \left( {\frac{{Ex_{il} }}{{X_{\hbox{max} } }}} \right)^{q} } \right)}^{{\frac{{w_{i} w_{j} }}{{1 - w_{i} }}}} } \right)^{{\frac{1}{p + q}}} ,\;\left( {\sum\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\frac{{w_{i} w_{j} }}{{1 - w_{i} }}En_{ij}^{2p} En_{il}^{2q} } } \right)^{{\frac{1}{2p + 2q}}} ,\;\left( {\sum\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\frac{{w_{i} w_{j} }}{{1 - w_{i} }}He_{ij}^{2p} He_{il}^{2q} } } \right)^{{\frac{1}{2p + 2q}}} ,} \right. \\ & \quad \left. {X_{\hbox{max} } \left( {1 - \left( {1 - \prod\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\left( {1 - \left( {1 - \frac{{ex_{ij} }}{{X_{\hbox{max} } }}} \right)^{p} \left( {1 - \frac{{ex_{il} }}{{X_{\hbox{max} } }}} \right)^{q} } \right)}^{{\frac{{w_{i} w_{j} }}{{1 - w_{i} }}}} } \right)^{{\frac{1}{p + q}}} } \right),\;\left( {\sum\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\frac{{w_{i} w_{j} }}{{1 - w_{i} }}en_{ij}^{2p} en_{il}^{2q} } } \right)^{{\frac{1}{2p + 2q}}} ,\;\left( {\sum\limits_{\begin{subarray}{l} l,j = 1 \\ l \ne j \end{subarray} }^{m} {\frac{{w_{i} w_{j} }}{{1 - w_{i} }}he_{ij}^{2p} he_{il}^{2q} } } \right)^{{\frac{1}{2p + 2q}}} } \right). \\ \end{aligned} $$

(11)

The aggregated results after using Eqs. (10) and (11) are still LINCs and satisfy the properties of reducibility, commutativity, idempotency, monotonicity and boundedness. The processes proving such findings are omitted here.