The second patent is an evolution in mobile marketing.It uses data gathered from the first patent or a similar program and creates mobile advertisements designed for an individual user.This is a brand new mobile advertising model，completely different from any mobile advertising method on the market today. While there are a lot of applications on the market for pairing and recommending products，these fail to pair up people (such as ad models or mobile game users). 

        According to analysis and statistics in the advertising field，most of the audience's first sense of attention is directed towards the model's face. In the case of mobile advertising，you have 3 seconds to hook people with your advertisement. The choice of the model and their fit with the demographic andsubsequent attraction to the model is one of the key elements in the success of an advertisement. If you choose the wrong model，or the audience does not like the chosen model，this will greatly damage the effectiveness of any advertisement.Excluding the use of famous stars，it is not easy to understand the aesthetic preferences of your demographic. Also，supermodels and big stars are expensive and often have conditions or restrictions (such as the corporate image or placement ...)，so many advertisers have difficulty finding a proper model for their ad.In addition mobile web ad layout is small and limited，and the role of the model is more conspicuous. While there may be no conscious recognition of a face in the image，if the audience has an innate attraction to the model's face the viewer will be more interested in the ad and subconsciously associate the image with a positive connotation.We believe the effectiveness of the advertising will be demonstratively better. 

A Review of The System and Method for Creating Individualized Mobile and Visual Advertisements Using Facial Recognition

By Dr. Waleem Alusa

Dr. Waleem Alusa Profile.pdf

        The innovation embodied in this patent involves inferring consumer preferences and taste from their choices of aesthetically pleasing models, and employing such inferences to supply them with targeted advertisements that are likely to meet their needs. This idea is novel and has a disruptive potential on the advertisement industry in positive ways. the patent dispenses with the traditional approach of inferring consumer tastes and preferences from aggregated or spatial data, and derives consumer preferences from actual sets of consumer sub-conscious decisions.

        The work-flow of the patent starts with presenting a set of aesthetically appealing visuals to a consumer, for them to choose a visual that is most appealing to them. The features that characterize the chosen visual is then recorded, and the process starts again. This process is repeated many times, with data continuously gathered about successive consumer choices. The data generated from consumer choices is then modeled with a machine learning algorithm for the purpose of predicting future consumer preferences. The predicted preferences informs the nature and type of advertisements or model offerings to be presented to the consumer in the future.

        A crucial element underlying the patent is the logistic regression method that is used to predict future consumer preferences. The method has been proven in many applications to be extremely useful in predicting binary outcomes.

        It is however limited in its ability to provide alternative options. In other words, the method can provide a determination as to the likelihood that a consumer would prefer an item. The model is limited to being able to answer a question such as "what is the likelihood that a consumer would prefer a model?" Perhaps, the patent can benefit significantly from methods that are able to answer questions such as "among a finite number of available models, which is the consumer likely to prefer, given his historical preferences inferred from past choices?" The statistical models used to answer this kind of questions are prevalent, and can be easily adopted and tailored to this patent.

        This method can be based on a machine learning algorithm that learns to recommend models that are similar to the ones that the consumer liked in the past. The similarities of models can be derived based on the features associated with the compared models.

        Notwithstanding, this patent is both innovative and novel, and has a positive potential high side, if properly implemented.