Competitive Intelligence (CI) provides critical decision support for the performance of a data-driven organization, identifying market opportunities, threats and risks. Artificial intelligence (AI) provides key tools and techniques to support CI in order to address an ever-changing market.
Artificial intelligence has the potential to completely revolutionize the data collection process. It allows to provide and analyze information promptly and deepen thousands of sources. It is also a more accurate process as a machine has a qualitatively higher data analysis capability than a manually conducted process.
Imetrics applies AI mechanisms to its tracking and analysis activities of intelligence data relating to competitors that allow it to combine the extracted data with intelligent processing algorithms to learn from patterns and characteristics of the data they analyze. Data derived from Imetrics’ ETL process (Extraction, Transformation, Load) is historicized in databases and then used as a source to feed different possible levels of data analysis through data mining functions and Artificial Intelligence algorithms (Linear regression, Decision Trees, Lasso, .., custom).
The potential of Artificial Intelligence extends from the initial phase of data collection by collecting and analyzing an ever-growing availability of relevant data sources. Not manually extracting information means spending less time researching information and making it easier to track updates and changes in real-time. The further selection of the most relevant information allows you to focus on insights that can be exploited by the business and which give meaning to a large amount of data available. The use of AI platforms (IBM Watson) varies the observation point from the “real” data to the probabilistic data, generating patterns that can better define all those variables that contribute to minimizing the risk rate of the decision-making process.
Imetrics proposes the framework adopted for an AI-enhanced Competitive Intelligence project with practical application to the Healthcare sector.
The Artificial Intelligence application was developed from a data set of different pharmaceutical brands for healthcare products on the market through online channels.
Products’ pricing was chosen as the main dependent variable to be studied according to the brand and their category of belonging.
The average price of each brand’s product, within the various categories, was calculated to exclude any possible impact generated by other variants of the product. It was calculated using different methods of data aggregation, assigning them the labels “Low“, “Medium” and “High” (price level), so that it would be possible to apply the chosen algorithms.
Algorithms were applied to the data obtained: a “supplementary tree structure classifier” whose accuracy (percentage ratio between correctly classified instances and total instances) equals 67.14%. It was possible to demonstrate the presence of a relationship between the price level of the products within the individual categories and their brand. A model of this kind is useful for dynamic predictions as well as for comparing the behavior of pharmaceutical companies and their brands concerning product pricing.
Specifically, the output of the applied AI model generated the following results:
The evaluation matrix of the statistical model used shows that 715 brand-category binomials, compared to the dataset, are correctly detected (Correctly Classified) at 14.3%, 87.5%, 5.7%, respectively in the Low, Medium, High market price ranges.
For all the Low, Medium, High market price ranges, the ROC Area is always over 0.5, indicating that the model used has a good degree of reliability (see next graph).
The ROC curve graphically represents the accuracy of the applied model, given that a curve with subtended area equal to 1 represents the “perfect” classifier while a ROC curve equal to the bisector of the graph with an area equal to 0.5 represents a completely random classifier. The ROC curve that the model generated has an area equal to 0.530. It can therefore be said that the model has good classification and no random ability.
Imetrics provides data and solutions for Competitive Intelligence
Imetrics is the partner of companies’ CI teams and provides AI-powered competitive intelligence expertise with transparent and innovative methodologies.