Annual Pharmaceutical Sales Estimates Using Patents

Pharmaceutical developments are often lengthy, requiring decades in certain cases, and entail enormous costs that may exceed $1 billion for a single product. Therefore, pharmaceutical development as well as marketing companies would benefit greatly from sales volume estimates of products at their early stages of development. Moreover, it would be desirable to estimate not only total sales but also annual volumes. In this work, we propose an artificial intelligence model to estimate future yearly pharmaceutical sales volumes. Accordingly, we use actual annual sales from pharmaceuticals and future sales forecasts provided by experts as data. For sales estimation, the attributes related to the pharmaceutical, such as manufacturer and therapy area, are used along with patents filed in the early stages of development. With regard to patents, attribute information such as the count of claims, inventors, and citing references are used along with text in the patent application for sales estimation. Experimental results show that the estimation accuracy improves as the number of words in the text used for estimation increases and by using the actual sales data from a few years after launch.