US8756174B2 | 2014-06-17 |
Claims Following is the claim for this invention:- 1. In this invention we have application which receives similar objects along with number of user likes for the object, the number of users who marked the object as favourite, number of comments made by different users for the object and classification of user response as positive, neutral or negative by doing sentiment analysis, number of users actually aware of the object via viewing or visiting the object (like viewing video objects or visiting objects like hotel), average rating per object per user, etc. For ranking such objects we would use machine learning which would require a training set. The training set should contain important and unimportant example objects. Now we have reduced our problem to a simple linear classification problem for which we can use several available machine learning algorithms like Support Vector Machine etc. Our classification problem at hand is to classify any new incoming object received by the application as important or unimportant. Suppose if we have N features (features of different objects are received by the application along with the object as explained above), the Support Vector Machine will train on the input training set we provide for the application and obtain a weight for each of these features. So we will have a weight vector of size N equal to the number of features. This weight signifies the importance of each attribute or feature in classifying the object as important or unimportant. Now using Support Vector Machine we will know which feature has more weight and give us better insight into which feature is the most important and we can rank all the important objects according to the value held by the feature carrying the highest weight. We can also give the user option to specify according to which feature they want to rank all the important objects they have received. The above novel technique of ranking important objects received by an application according to a feature or attribute of highest significance is the claim for this invention. |
In this invention we have an application which receives similar objects along with number of user likes for the object, the number of users who marked the object as favourite, number of comments made by different users for the object and classification of user response as positive, neutral or negative by doing sentiment analysis, number of users actually aware of the object via viewing or visiting the object (like viewing video objects or visiting objects like hotel), average rating per object per user, etc. For ranking such objects we would use machine learning which would require a training set. The training set should contain important and unimportant example objects. Now we have reduced our problem to a simple linear classification problem for which we can use several available machine learning algorithms like Support Vector Machine etc. Our classification problem at hand is to classify any new incoming object received by the application as important or unimportant. Suppose if we have N features (features of different objects are received by the application along with the object as explained above), the Support Vector Machine will train on the input training set we provide for the application and obtain a weight for each of these features. So we will have a weight vector of size N equal to the number of features. This weight signifies the importance of each attribute or feature in classifying the object as important or unimportant. Now using Support Vector Machine we will know which feature has more weight and give us better insight into which feature is the most important and we can rank all the important objects according to the value held by the feature carrying the highest weight. We can also give the user option to specify according to which feature they want to rank
all the important objects they have received.
Next Patent: BASEBAND FREQUENCY SELECTIVE MAGNITUDE AND PHASE ADJUSTMENT FOR WIDEBAND DOHERTY POWER AMPLIFIER