Research On Mining Method Of Mine Probabilistic Stream Data

Mine Probabilistic Stream Data means that the data in coal mine safety monitor system is satisfied with probability data stream model. Research on mining method of Mine Probabilistic Stream Data has the goal of predicting the mine disaster, researching mining system architecture of Mine Probability Stream Data, and the lightweight minging method under the restriction of the low time and space complexity, incremental maintenance, adaptive and description of probability . To overcome these limitations, the main research work includes:1. Under analizing the goals and the characteristics of the mine Probabilistic stream data mining system, Referring to the system architecture of centralized and streamed, aiming at the forecast of mine disaster, construction the mining system is fit for mine probabilistic data stream, including system model, architecture and hierarchical structure.2. Researched the pattern representaion problem of mine probabilistic stream data.Referencing the Piecewise Linear Fitting of traditional time series, this paper advanced a Fitting Point based Piecewise Linear Fitting method, shorted as FP_PLF method. FP_PLF method does not rely on the factor of whole length of time series and expert's domain knowledge, also satifies the mining algorithm requirements that the low time-space complexity, incremental maintenance, adaptive and description of probability .3. Researched the problem on finding the shape abnomal of the beginning of the mine disaster, this paper put forward the conception of pattern abnomal detection for mine probabilistic stream data .Based on the similarity distance between two probilistic stream, advancing a Probabilistic Similarity Distance based Pattern Abnomal Detection method, shorted as PSD_PAD method.The PSD_PAD method has good effections on context uncorrelated abnomal and has low time-space complexity in the condition that the window size smaller than 30.4. Researched the problem of transmission of pattern abnomal probability in mine probabilistic stream data mining system.This paper, we put forward a Pattern Abnomal Dirrected Diffusion algorithm, shorted as PADD algorithm. The PADD alogrithm has a higher network delay,but lower missed report rate than traditional WSN ruting protocols and implements the abnomal probability completed to the server, reduce network energy waste, postpone the life of data mining system. Provide hardware protection for ultimate mining goal of the system.5. Researched the problem of mine disaster forcasting based on the mine probabilistic stream data.In this paper, based on exception test method of trend analysis, proposed a Trend Analyze based Disaster Abnomal Detection method ,shorted as TA_DAD method. The TA_DAD method make use of monitoring gas data in the normal times to establish the forecasting model during the hazards period to solve the difficulties of data acquisition; solved the problem of predicting in nonlinear systems and the nonlinear method complexity and high complexity of the time and space problem in use of linear prediction method, combination with exception test method, you can quickly and accurately to achieve the mine disaster forecasts.6. By the construction of prototype system of mine probabilistic stream data, this paper realized the model representation, pattern anomaly detection, anomaly information transfer and disaster anomaly detection of the mine probabilistic stream data mining system, which proved the feasibility and effectiveness of data mining method of the coal probability data stream, and put forward new ideas and new exploration for mine disaster prediction.