A Study Of Product Innovation Strategy By Computational Experiment Based On Online Data Analysis

With the deepening of the network economy, e-commerce generated huge number of online data. Companies can get more depth and breadth information about the product and the consumer by these data. This information is the power of enterprise product innovation and also an important factor which impact the effectiveness of product innovation strategy. Based on the empirical analysis of Amazon’s online data using the methods of SVM and Apriori, this paper found out the attention and satisfaction of product features and consumers’behavior preferences. Then we used the method of multi-agent model within computational experiment to characterize the interaction rules of the consumer agent, businesses agent and the market environment. selecting electronic products as an example, we study the effectiveness and influence factors of three products innovation strategies which includes technology-preferred, market-preferred and balanced.Firstly, the paper built the features extraction and views recognition model and reviews emotion classification model with method of SVM. With the model of features extraction and views recognition we found that price, performance, appearance, After-sales service, battery life, technological level, experience, storage, systems, accessories, radiating are the most frequently commented product features. The satisfaction of various product features is different. Affected by products’innovation strategies the attention and satisfaction of features are not directly related. Though the trend of attention on features in last ten years, we found the consumers’trend of preference on features. With the model of reviews emotion classification, the paper found that consumer prefer to read reviews with negative emotions and they believe those reviews are more useful. With the mothed of Apriori we built the model of features correlation analysis. With this model we found that some features are most likely being commented together. And this relevance would be called transition probability matrix represented by confidence between features commented in reviews.Based on the conclusion of the empirical analysis, this paper used the method of computational experiment and the model of multi-agent to design the consumer agent, the business agent, also the market environment. In the process of designing the consumer agent, the paper built five models as follow:the model of consumer characteristics, the model of consumers’preferences evolution on various product features, the Bayesian model of consumers’belief updating, the maximum utility model of consumers’purchasing decision, also the model of evaluation about product. In the process of designing the business agent, the paper built another four models as follow: the MNK models of multi-characteristics complex product, the model of product innovation strategy, the Logit model of business’producing and pricing, also the model of calculating cumulative profit. Based on the design of market agent, the paper completed the design and implementation of the entire e-commerce supply chain systems.In the process of computational experiments study, this paper selected electronics as an example to analyze the effectiveness and influences of product innovation strategy. We found that the performance of products innovative strategy is related to product lifecycle. At the beginning of lifecycle balanced strategy is superior, after that technology-preferred strategy is better. The market structure, external information, and consumers’learning preferences will influence the performance of innovation strategy. Firstly, the number of enterprises carrying out different innovative strategies will affect the performance of each strategy. Secondly, availability of other enterprises’online data is an important factor which influences the market-preferred strategy’s performance, also have an indirect influence on technology-preferred strategy. Lastly, the bigger is the proportion of negative reviews which have been learned by consumers, the better performance is the performance of balanced strategy and the worse is performance of technology-preferred strategy.