An ant colony optimization-based classifier system for bacterial growth

摘要: Motivation. In predictive microbiology, identification of different combination environmental factors (such as temperature, water activity, pH), which lead to growth/ no–growth microorganism, is a problem potential importance. Ant colony optimization (ACO) one the most recently developed nature–inspired metaheuristic techniques, based on foraging behavior real life ants and has already exhibited superior performance in solving combinatorial problems. This work explores search capabilities this for learning classification rules bacterial growth/no growth data pertaining pathogenic Escherichia coli R31 affected by temperature activity. The discovered thus can be used verify whether any activity belong either or microorganism. Method. ant algorithm works iteratively follows: At iteration level, software construct using available heuristic information dynamically evolved pheromone trails. A rule that highest prediction quality said rule, represents extracted from database. Examples correctly covered are removed training set, another started. Guided modified matrix, agents build improved process repeated many iterations necessary find covering almost all cases set. Results. ACO classifier system utilized several datasets its compared with other well known algorithms terms average accuracy attained 10–fold cross validation. results obtained compare very favorably classifiers. Additionally, discovery dataset growth/no–growth, C4.5 respect simplicity rules. both these indices compares C4.5. Conclusions. sets indicate competitive considered useful tool knowledge given