Reserch On Rule Discovery In Time Series

Time series data is a series of observations arranged in chronological order.Time series have the characteristics that the amount of data increases with the passage of time.Extracting the implicit information in the data is of great significance for finding the interaction between things.Data mining provides a good research idea and method for time series analysis.However,since one important characteristic of time series data is that each discrete independent point is not meaningful,data as a whole is meaningful.Because the idea of association rules based on discrete numbers is directly used,the traditional study of time series rules discovery algorithm is stagnant.Therefore,this paper proposes a rule discovery algorithm based on the shape of time series.Based on the time series motif discovery algorithm,the algorithm improves the traditional rule form and extends it to the rule form that causes consequence to occur under two antecedents combination constraints.At the same time,a measure of whether the algorithm can discover the internal structure or internal value of the time series data is proposed.Finally,through comparative experiments,the accuracy and rationality of the algorithm is confirmed.For the reality that different time series are related in the relevant time,a multi-dimensional time series rule discovery algorithm based on motif is proposed to solve the problem of information loss and dependency parameters caused by the discretization of multi-dimensional time series.Firstly,this algorithm discovers the association rule items of time series in each dimension independently.Secondly,it uses the time range to associate the rule items of different dimensions together.Finally,the rules of the entire multi-dimensional time series can be found by discovering the correlation between the rule items of different dimensions.At the end of this study,the experimental results show that the proposed algorithm has the advantages of good scalability and solve the problem of high time consumption and large influence by random walk data.