A mass assignment method for prototype induction

Label prototypes for modelling with words

“…This is an alternative to the approach described in [1] and [2] where prototypes are tuples of fuzzy sets on labels.…”

Fuzzy prototype model and semantic distance

“…It is more robust, accurate and describable model to solve various types of problems in the real world. However, the previous model proposed in Baldwin's paper in 1999 [1] is only limited to infer a new data example to a certain fuzzy prototype. Therefore, we propose an improved fuzzy prototype model, which is able to make an inference for a new example giving a membership degree for each fuzzy prototype, and which still keeps the function of the previous fuzzy prototype model.…”

“…The first method is to classify a new data example as a certain fuzzy prototype, and the second is to give the new example a membership of belonging to each fuzzy prototype. Compared with the previous model [1] only available for the first method, the improved fuzzy prototype model introduced in Section 2 is suitable for both methods.…”

Classification and query evaluation using modelling with words

“…These results illustrate that now the breadth first search is obtaining the highest classification accuracy with an improvement from Naïve Bayes of 3.3%. Using this database a direct comparison can be made with results from Baldwin et al [1], who split the database into two equal training and test sets of 107 examples. Here using the improvement measure with a threshold of 0.895 and a maximum grouping of 4 attributes, a classification accuracy of 71.03% on the test set and 92.52% on the training set can be obtained by applying breadth search.…”

“…Here using the improvement measure with a threshold of 0.895 and a maximum grouping of 4 attributes, a classification accuracy of 71.03% on the test set and 92.52% on the training set can be obtained by applying breadth search. This compares well with the results of 71% on the test set using a mass assignment prototype method [1] and a test set accuracy of 68% using a mass assignment ID3 system [2].…”