巫啓豪
巫啓豪
馬斯洛需求層級理論為基之產品定位分析機制研發
Development of a Product Positioning Analysis Mechanism based on Maslow's Hierarchy of Needs Theory
意見探勘、馬斯洛需求層級理論、產品定位、凝聚式階層分群演算法、倒傳遞類神經網路
Opinion Mining、Maslow’s Hierarchy of Needs Theory、Product Positioning、Agglomerative Hierarchical Clustering、Back Propagation Neural Network
行銷為提供產品滿足消費者需求與慾望之過程，企業則藉由行銷活動創造顧客與留住顧客。而產品定位分析是企業擬定有效行銷策略的重要步驟之ㄧ，其目的為得知消費者對於產品之特徵的重視程度，使企業能夠塑造產品的鮮明個性或特色。現下，隨著各種網路媒體的蓬勃發展，消費者無不透過這些媒體分享其對產品的意見及使用心得等評論文章。這些文章往往透露出消費者對於產品特徵的重視及需求。本研究試圖以馬斯洛需求層級理論(Hierarchy of needs theory)為基礎，發展一個產品定位分析(product positioning)的機制，透過關聯法則(Association Rule)之觀念以及機器學習中凝聚式階層分群(Agglomerative Hierarchical Clustering, AHC)、倒傳遞類神經網路(Back-Propagation Neural Network, BPN)等方法分析網路媒體中消費者對於產品之評論，探勘出目標客群對於特定產品所重視的產品特徵及各特徵對應至馬斯洛需求理論之層級，以建立一消費者的產品馬斯洛模型。此模型將可提供企業審視產品在目標市場之定位，並作為產品改良或開發的決策支援。
This study develop a product positioning analysis mechanism based on Maslow's hierarchy of needs theory, so that enterprises can understand the level of needs in target audiences (TA) psychological dimension in low-cost and automated way. The main contributions of our paper are three fold: (1)We designs a product’s Maslow model that provides information of product positioning in target market to enterprises, so as to support refining product positioning strategies, improving features of products, and making decisions. (2)We propose a FOPE Algorithm which extend the method which proposed by Liu et al. (2013), which extracts features and opinions of product from online word-of-mouth articles (reviews) written in Chinese. (3) We propose an Aspect Induction Algorithm that summarized the product features which have similar meaning to a cluster called aspect by AHC (agglomerative hierarchical clustering). (4)We designs a Maslow levels evaluation method that uses Back-Propagation Neural Network (BPN) to map each product features which consumer emphasized to five levels of Maslow's hierarchy of needs. Based on these methods, we conduct an empirical study on smart phones and butter Shortcake and demonstrate the effectiveness of the proposed methods.