Locust optimization-based situation analysis method and system for LSSVM-ARIMA model, and storage medium

Abstract

The invention provides a situation analysis method and a system of an LSSVM-ARIMA model based on locust optimization and a storage medium, belongs to the field of machine learning and data mining, andis characterized by comprising the following steps: (1) randomly initializing an initial position of a locust group; (2) determining a target function; (3) updating the position; (4) repeating the steps (1) and (2), and outputting c and sigma; (5) establishing an LSSVM model and an ARIMA model; calculating a prediction result y1 (t); (6) determining a low-frequency component Aj (t) and a high-frequency component Dj (t); (7) obtaining a first prediction result y1 (t); (8) calculating a prediction result y2 (t); and (9) fitting the prediction results y1 (t) and y2 (t) to obtain a final situation result y (t). According to the method, a mode of combining the LSSVM and the ARIMA is adopted, and the locust optimization algorithm is utilized to realize parameter optimization on the time-varyingsituation model. Experimental results show that the enterprise safety production situation prediction method established by the invention is effective, and a reliable method is provided for enterprise safety production management situation analysis.

Description

technical field [0001] The invention relates to the technical field of machine learning and data mining, in particular to a situation analysis method, system and storage medium based on a locust-optimized LSSVM-ARIMA model. Background technique [0002] At present, for the situation prediction problem, most models can fit the original data to a high degree, but the generalization ability of the model is poor. For the situation prediction problem, since the data is a time series and changes in real time, these models often show a good fitting effect on historical data, but in some emerging data, the predictive ability will be greatly reduced. Although the neural network has good generalization ability and memory ability, the convergence speed is too slow during the model training process, resulting in too long training time, which cannot meet the timeliness requirements of numerical prediction. Taking the recurrent neural network RNN ​​as an example, the calculation result o...

AI Technical Summary

Problems solved by technology

However, if long-term memory is required, the calculation of the current hidden state needs to be linked to the calculation of the previous n times, which makes the calculation amount increase exponentially, resulting in a significant increase in the training time of the model

Similarly, when the BP neural network uses the error backpropagation method to adjust the weights between neurons in each layer, it often needs to go through multiple iterations, which increases the training time of the model and easily falls into a local optimum during the training process.

Images