Towards Contiguous Sequences in Uncertain Data

In data mining, high-utility sequential pattern mining (HUSPM) focuses more on the specific values of items than on their frequency, making it more practical in real-life scenarios. HUSPM with the contiguous constraint can be used to solve some applications requiring the sequence elements to occur consecutively. Due to device, environment, privacy issues, and other factors, the data is often not accurate, and traditional algorithms for mining high utility continuous sequence patterns (HUCSPs) do not perform well in handling uncertain data. To address this challenge, this paper presents a new algorithm named uncertain utility-driven contiguous pattern mining (UUCPM), which can discover HUCSPs efficiently and correctly. The algorithm is designed to obtain results from sequence data with uncertain probabilities set on the item level. Two tighter upper bounds on utility and corresponding pruning strategies are also proposed, which can effectively process and reduce the number of candidate patterns generated during pattern mining, thereby improving the performance of the mining process. Through extensive experiments, the proposed UUCPM algorithm has been verified for accuracy and performance, demonstrating its advanced properties.