An extended MDP-PROMETHEE algorithm for MAGDM based on social media data mining under probabilistic linguistic environment

With the popularity of the Internet, more and more people focus on the evaluation information of users on social platforms. When new users purchase some products, the decision-making between the alternatives will be greatly influenced by the evaluation information of large groups of users. Fuzzy linguistic term set is often used for users to express their evaluation information of products. The web crawler technique is used to obtain the online comments by users, then the public-level weight of alternatives on each attribute is calculated by TF-IDF and Word2Vec. Meanwhile, the experts' evaluation information about alternatives is obtained by some questionnaires and calculate the expert-level weight of alternatives on different attribute. The public-level weight and expert-level weight are fused to obtain the final weight distribution of each alternative on different attribute. The online comments of large groups of users and the evaluation information of experts about alternatives are integrated to ensure the scientific and completeness of evaluation information. Under probabilistic linguistic term sets, a new generalized hybrid distance is proposed to improve the traditional Euclidean distance, Hamming distance and Hellinger distance. The variance parameter based on Gaussian criterion is replaced by the range parameter, and Gaussian distribution is improved. Then an improved preference function is given based on the new generalized hybrid distance. Under the discussion above, a novel MDP-PROMETHEE algorithm is proposed. By the case study on six cars purchasing decision making in E-commerce, MDP-PROMETHEE is demonstrated to be scientific and valid. Meanwhile through the comparative analysis with PROMETHEE and TOPSIS, MDP-PROMETHEE is verified to be more reasonable and superior.