Aggregation pheromone metaphor for semi-supervised classification

8 This article presents a novel ‘self-training’ based semi-supervised classifica9 tion algorithm using the property of aggregation pheromone found in real 10 ants. The proposed method has no assumption regarding the data distribu11 tion and is free from parameters to be set by the user. It can also capture 12 arbitrary shapes of the classes. The proposed algorithm is evaluated with 13 a number of synthetic as well as real life benchmark data sets in terms of 14 accuracy, macro and micro averaged F1 measures. Results are compared 15 with two supervised and three semi-supervised classification techniques and 16 are statistically validated using paired t-test. Experimental results show the 17 potentiality of the proposed algorithm. 18