The rule-based classification algorithm now includes a dynamic threshold value. The significance of threshold values, classification algorithms that use threshold values, issues with setting static threshold values and dynamic threshold values, and rule generation. However, for some datasets, it is necessary to have a large number of rules, and for those datasets, a greater number of rules should be extracted, which affects reduction of generalisation and makes the system less transparent. Logic rules framed for small datasets have minimum number of rules that interns are very easy to understand and compare. One of the best, most flexible solutions is to frame fuzzy logic rules; however, fuzzy rule support is less for datasets with symbolic and nominal attributes. These alternative rules extraction systems are driven by similarity-based learning and are based on prototype rules, presented threshold rules algorithm, which uses a small number of highly precise ordered rules to extract threshold rules from data.