307 results about "Exponential smoothing" patented technology

A computer implemented system and methods for forecasting the future values of a time series with multiple seasonal cycles are described. The forecasting method of the present invention include 1) two multi-seasonal exponential smoothing methods, 2) the seasonal sub-series method splitting a time series into multiple sub-series, and 3) the aggregate-to-period forecasting method.

The invention discloses a multi-time scale forecasting method for a road traffic running situation. Highway traffic parameters in different time scales are analyzed according to the running time-space characteristics of highway traffic flow; the highway road traffic running situations in different time scales are forecast by an exponential smoothing algorithm, a weighted average algorithm and a Kalman filtering method respectively; a highway road traffic running situation evaluation index system and a multi-time scale highway traffic flow running situation forecasting technology are constructed to implement the conversion from experience guide to science guide for the highway running management and the preliminary conversion from passive management to active management. Therefore, the running efficiency of a road traffic running situation forecasting system can be increased effectively, the running cost of the system is reduced, the coordination degree between road traffic guidance and management can be improved obviously, and an optimal policy is provided for improving a traffic management and control measure and planning a travel plan for a road traffic manager and a user to a large extent.

An anomaly detection engine monitors network traffic to detect orders placed by users from an electronic catalog of items, aggregates data about the detected orders by time period, and analyzes the aggregated data to detect anomalies in activity levels associated with specific items in the catalog. To detect whether an anomaly exists in the activity data associated with a given item, a forecasting algorithm, such as an exponential smoothing algorithm, is used to generate an expected order volume for a current time period, and the expected order volume is compared to an actual order volume. Other criteria may also be taken into consideration. If an anomaly is detected, such as a sharp increase in the item's order volume, the anomaly detection engine generates an alert message to notify a catalog administrator, who may then determine whether the anomaly is attributable to an erroneous item description in the catalog.

The invention discloses a memory caching method oriented to range querying on Hadoop. The memory caching method oriented to the range querying on the Hadoop comprises the following steps that (1) an index is established on querying attributes of Hadoop mass data and is stored on an Hbase; (2) a memory is established on index data of the Hbase to conduct fragment caching, the frequently-accessed index data are selected and stored in the memory, data fragments are fragmented in an initial stage by adopting a fixed length equal dividing method, and the mass data fragments are organized by adopting a skiplist; (3) hit data are queried and recorded according to the data, and the heat of the data fragments is measured by adopting an exponential smoothing method; (4) a memory cache is updated. The memory caching method oriented to the range querying on the Hadoop has the advantages that the structure of combining the skiplist and a collection is adopted, the dynamic adjustment of the fragment boundary of the collection is supported on the structure, the data fragments are made to be adaptive to querying demands, the querying cache hit rate of hot data fragments is improved, the overhead of a querying accessed disk is lowered, and thus the performance of the range querying is improved greatly.

A high-speed rail train driving method based on the section profile passenger flow includes the following steps that high-speed rail station information is obtained through a high-speed rail network operating system firstly, and a high-speed rail network is divided into a plurality of sections; the stop modes of high-speed rail trains are determined according to multiple grades; the O-D (origin-destination) demands of each section are classified according to the stop modes; the historical data about the actual O-D demands of each section are obtained, and the time varying demand of each type of O-D in each section is obtained through a linear second exponential smoothing prediction method; on the basis, the driving scheme unit of each section is made according to the sequence of the trains from the high grade to the low grade; all the driving scheme units are spliced and combined to form an initial train driving scheme; and finally, the scheme is adjusted to meet various restraint conditions and optimized through a simulation annealing algorithm, and each high-speed rail train is driven according to the optimization result. The high-speed rail train driving method based on the section profile passenger flow takes the benefits of the high-speed rail system and the passenger demands into account, and can improve the running benefits of the high-speed rail system.

The invention provides an automatic detection method of a traffic incident on a highway, which is used for automatically detecting a traffic incident on a highway in the case of deficient processing data. The automatic detection method comprises the following steps of: acquiring real-time traffic flow information of a detection area according to a set time interval by an induction coil or video detection equipment through a detection system; preprocessing (including normalization, discretization and the like) the acquired real-time traffic flow information; inputting the preprocessed data into a detection algorithm based on a Navie Bayesian classifier so as to obtain posterior probability of the incident; and exponentially smoothing the obtained posterior probability, sending an alarm and informing a traffic management department to take corresponding measures to handle the incident if the smoothed posterior probability is higher than a threshold value, and otherwise proceeding to acquire the data and carrying out next judgment. The automatic detection method provided by the invention has the advantages of high detection rate, low false alarm rate and short detection time, and the deficient data can be processed by the method, thus the method can be widely applied to a highway management system.

The invention discloses a medium and long term demand forecasting method for tendency and periodicity commodities. The method is suitable for medium and long term forecasting of the commodities which have periodicity, are influenced by trend variation and are sensitive to seasons. On the basis of exponential smoothing, linear trend, seasonal variation and random variation time series are decomposed, full consideration is given to influences of the trend variation and period variation on commodity transaction data, commodity transaction data according with a forecast model are selected according to a commodity data selection principle, corresponding smoothing parameters are determined, various indexes in the forecast model are respectively calculated through a quantitative method, and a predicted value of the commodities in a certain period in the future is obtained. Therefore, scientific judgments are provided for production, manufacturing, storage, marketing and the like of various enterprises in commodity supply chains.

The invention belongs to the technical field of network information prediction, and discloses a time sequence prediction method based on a GRU neural network. The method comprises the following steps:collecting original data needing to be predicted; carrying out data preprocessing on the collected original data; performing standardization processing on the preprocessed data; performing dimensionraising processing on the original time series data form by using the code; training the input data by using a GRU neural network to obtain a trained time sequence prediction model, and storing the trained time sequence prediction model; predicting the time series data by using a GRU-SES model to obtain a preliminary prediction value; performing secondary exponential smoothing processing on the obtained preliminary prediction data to obtain a final prediction data value; output of prediction results. The prediction method provided by the invention improves the precision of time sequence prediction, and is of great significance to time sequence analysis in industrial production or actual life.

The invention discloses a realization method for path navigation of an intelligent navigator. The realization method comprises the following steps that: (1), a navigation data transmission module sends vehicle driving data to a navigation data storage module at fixed time; (2), a navigation data analyzing module carries out statistics on driving data of a single vehicle and establishes a social map and a vehicle driving habit record; (3), the navigation data analyzing module analyzes driving data of multiple vehicles at all road sections and establishes an average speed oscillograph of all the road sections, and a road condition report and a trend report are formed by using a second exponential smoothing method; (4), the navigation data analyzing module matches the vehicle driving data with the driving habit record and predicts a vehicle driving destination and a driving route; and (5), according to the road condition report and the trend report, the navigation data analyzing module predicts real-time road condition information and real-time path recommendation information, and sends the information to the navigation data transmission module. According to the invention, characteristics of high practicability, high real-time performance and large information content and the like; and the real-time performance and the personalized service capability of the intelligent navigator are improved.

The invention discloses a method for calculating the trust values of wireless Mesh network nodes. The method comprises direct trust value calculation, indirect trust value calculation and comprehensive trust value calculation, wherein the direct trust value calculation is carried out for acquiring the interaction times of different time slices among the nodes and establishing a time sequence according to the obtained data at first, and then predicating the interaction times of the next time slice among the nodes by virtue of three times of an exponential smoothing method, and taking the relative errors of the interaction times predicated values and the actual values as the direct trust values of the nodes; the calculation formula of indirect trust values is obtained in a multi-path trust recommendation mode; comprehensive trust values are obtained by virtue of integrated computation for the direct trust values and the indirect trust values. The invention provides a method for calculating the trust values of the nodes, according to the specific condition of a network, adaptive smoothing factor alpha, credibility threshold value Phi, the value of a direct trust value weight beta are selected, the time attenuation characteristic and objectivity of the trust values are guaranteed, the credibility of the nodes is objectively and accurately described, the computation complexity is low, and the method is suitable for a wireless Mesh network.

A computerized method of optimizing audio quality in a voice stream between a sender and a receiver VoIP applications, comprising: defining by the receiver time intervals; determining by the receiver at the end of each time interval whether congestion exists, by calculating (i) one-way-delay and (ii) trend, using double-exponential smoothing; estimating by the receiver a bandwidth available to the sender based on said calculation; sending said estimated bandwidth by the receiver to the sender; and using by the sender said bandwidth estimate as maximum allowed send rate.

The present disclosure generally relates to a capacitance sensing apparatus equipped with self-calibrating capacity and method of use thereof. The disclosure contemplates the determination using a secondary means of precise fluid levels according to five possible embodiments, and the use of the determined fluid level to recalibrate the capacitance sensing apparatus along its continuous analog level, namely, a variation of the thickness of the insulation of a capacitance sensing apparatus, the variation of the surface geometry of the capacitance sensing apparatus, the use of a dual-probe sensor including a probe with a varied surface geometry, the use of an electromagnetic sensor adjoining the capacitance sensor, and the variation of the electromechanical sensor to serve as a capacitance sensing apparatus. The disclosure also contemplates methods for using the sensing apparatus previously disclosed to measure a fluid level using a self-calibrating capacitance sensing apparatus. Finally, the present disclosure contemplates the use of an improved mathematical method associated with a variability measurement, such as an exponential smoothing method, to determining locally discrete changes in the variability measurement of the capacitance in order to determine a fixed fluid level.

The invention relates to a data processing method and discloses a method for processing abnormal data of a real-time data acquisition system in real time. The method comprises the steps of (1) initializing sample data and selecting an even number of normally operating sample data; (2) adopting 1/2 of the sample data to act as the moving step by using a single exponential smoothing method, and predicting the latter half part of the sample data by using a single exponential smoothing recurrence method; (3) the residual of a prediction result is calculated according to a prediction value and a measured value of the latter half part; (4) carrying out anomaly analysis on the residual sequence according to a Pauta criterion to confirm whether the measured value is abnormal data or not; and (5) replacing the measured value with the prediction value if the measured value is abnormal data. The method disclosed by the invention is mainly advantageous in that a prediction algorithm coefficient is adjusted in an adaptive mode, the error is analyzed by adopting a mobile exponential smoothing method, and the anomaly judging method better conforms to use conditions of the Pauta criterion, thereby improving the accuracy in judgment for the abnormal data, and preventing false judgment and missing judgment to a certain degree.

The invention discloses a sinus heart rate turbulence trend detection method based on piecewise linearization. The piecewise linearization is adopted for specifically analyzing whether the variation trend of sinus heart rate accelerates first and then decelerates after premature ventricular contractions or not. The method includes the steps: (1) electrocardiosignal preprocessing; (2) self-learning process for first 10 seconds; (3) HRT (heart rate turbulence) sample collection; (4) piecewise trend analysis; and (5) turbulence trend representation based on a cloud model. The variation trend of the sinus heart rate at the RR interval is detected by means of piecewise linearization, and the turbulence trend is further represented by natural language through the cloud model. By the aid of an MATLAB (matrix laboratory) simulation tool, signals in an MIT-BIH heart beat irregularity database are selected for verification, and the variation trend of the sinus heart rate after single-time ventricular premature beat can be detected correctly. In addition, using the exponential smoothing method to predicate QRS complex occurrence positions to facilitate detection of QRS complex, and using a template for judging sinus heart beat is simple to implement and suitable for real-time treatment.

The invention discloses a power load short-term forecasting method, a model, a device and a system. Receiving influencing factor data, wherein the influencing factor data comprises air temperature data, holiday data and industry data, and quantifying the influencing factor by a quantization method corresponding to the load data; Wavelet decomposition is used to process the load data, and four sequences are obtained by de-noising and feature extraction. The correlation analysis is carried out with the data of influencing factors, and the set of correlation pairs of load characteristics and influencing factors is obtained. The four sequences obtained from the load data are preliminarily predicted by three exponential smoothing algorithms. The set of correlations between load characteristicsand influencing factors is predicted by ARIMA-GARCH method so that adjustment factors are obtained, and the preliminary prediction results and adjustment factors are input into long-term and short-term memory network, so final prediction results are obtained.

The invention discloses a method for analyzing a running trend of an electric power communication transmission network. The method is characterized by comprising the following steps: (1) acquiring basic data, and providing dynamic parameters and static parameters for a trend analysis model; (2) performing mutation analysis on performance; judging whether the acquired dynamic parameters are out of limit or not based on a performance parameter early warning threshold value; directly generating a performance alarm if the acquired dynamic parameters are out of limit; (3) performing trend analysis when the performance is steadily changed; constructing an error rate and optical power analysis model according to dynamically acquired performance parameters by adopting secondary dynamic exponential smoothing; obtaining a trend analysis weight through a steepest descent iteration method by combining the dynamic parameters and historical data, and further, constructing a linear prediction equation; (4) performing early warning computation, and outputting a performance prediction result. The method disclosed by the invention is applied to a communication management system, and therefore, a powerful data support is provided for the whole-process intelligent computational analysis from data acquisition, trend prediction, early warning analysis and a maintenance strategy.

The invention discloses a lithium ion battery remaining useful life prediction method. The method comprises the following steps that state parameter change data of a battery model is obtained by applying a particle filtering algorithm, the data are imported into an exponential smoothing prediction model (ES) so as to obtain a state parameter prediction value, then the state parameter prediction value is brought into an observation equation to obtain a capacity observation prediction value, and finally, the observation prediction value is fed back to the particle filter to predict the remaininguseful life (RUL) of a battery. According to the method, the ES-PF prediction model can solve the problem that state parameters cannot be updated in the prediction state by adopting the particle filtering algorithm, so that the prediction error is increased along with the change of the prediction period, and the prediction precision of the particle filtering algorithm is effectively improved.

The invention discloses a drug retail industry drug sales prediction system, and belongs to the technical field of big data and artificial intelligence. The system comprises a data acquisition and preprocessing unit, a historical data analysis unit and a prediction unit, and further provides a drug sales prediction method in the drug retail industry, and the method comprises the steps of data acquisition and preprocessing; Historical data analysis and data normalization processing; Data prediction: respectively establishing a cubic exponential smoothing prediction model, a grey prediction model and an LSTM prediction model, and combining the cubic exponential smoothing prediction model, the grey prediction model and the LSTM prediction model to form a BP neural network combination prediction model; Finally, a combined prediction result based on the BP neural network is obtained, so that a prediction model is rapidly and accurately established according to historical sales data, sales data of the pharmacy drugs in a future period of time is predicted, and then data support is provided for warehouse drug storage so as to achieve the purpose of meeting drug supply and demand balance.

The invention relates to a real-time detection and separation method of multiple moving objects by a through-the-wall radar. The method comprises: a complex background signal is rejected by using a clutter suppression algorithm based on an exponential smoothing filter and information of moving human objects can be kept rapidly and accurately; with an envelope detection algorithm based on Hilbert transform, all moving objects are arranged at different envelope peaks; and all moving objects can be separated at different areas rapidly according to an algorithm based on combination of a fixed threshold value and an adaptive threshold value. According to the invention, the method which is simple has the high execution efficiency and is suitable for detection on multiple moving human objects by a portable through-the-wall radar; a powerful guarantee can be provided for accurate positioning of all moving objects by the through-the-wall radar; and the working efficiency of the through-the-wall radar can be improved effectively.

The invention discloses a power distribution network operation efficiency evaluation method and system based on big data mining, and relates to the field of the power distribution network monitoring and control. The method comprises the following steps: collecting a power distribution network operation efficiency evaluation parameter, and establishing a power distribution network operation efficiency evaluation parameter database; constructing at least a multiple linear regression model, an exponential smoothing model, a first-order/second-order self-adaptive combination model and a neural network parameter optimization model; and evaluating and/or predicting the power distribution network operation efficiency by using the multiple linear regression model, the exponential smoothing model,the first-order/second-order self-adaptive combination model and the neural network parameter optimization model. By mining the relativity big data of the power distribution network, the distributionnetwork equipment and system operation efficiency, the coordination and the equipment equilibrium degree, the public power grid capital, and the user dedicated capital efficiency are analyzed according to different power supply regions and function zone types, and the influence factors with low efficiency are mined, thereby forming an efficient tool for distribution network efficiency monitoring,and a certain tendency pre-judgment and estimation function is provided.

The invention discloses a method for predicting the wind speed of a wind power plant at a short term. The method includes the following steps that wind speed data in a period of time are acquired and reduced into a time sequence for analytical prediction; a carpet traversal search method is adopted, and based on a dynamic cubic exponential smoothing prediction method, dynamic smoothing coefficients are determined according to the criterion of the minimum error sum of squares; the determined dynamic smoothing coefficients and the dynamic cubic exponential smoothing method are utilized to conduct one-step or multi-step prediction; the rest is conducted in the same way, new historical data are acquired, the smoothing coefficients are updated, and the later wind speed prediction continues to be conducted. According to the method for predicting the wind speed of the wind power plant at the short term, the characteristics of local historical wind speed are comprehensively considered, and the prediction effect is good.

The invention discloses a method and system for forecasting traffic flow data based on exponential smoothing. The method comprises the steps of conducting pre-processing, wherein collected traffic data are pre-processed; correcting errors, wherein filtering is conducted on the pre-processed traffic data according to the exponential smoothing method, reverse recursion is conducted on a filtering result, and therefore the errors in the filtering result are eliminated; conducting forecasting, wherein forecasting is conducted on the traffic data without errors through a cubic exponential smoothing forecasting model, and therefore traffic flow forecasting information is obtained. According to the method and system for forecasting the traffic flow data based on exponential smoothing, the traffic data which are collected in real time are pre-processed, so that abnormity in measured data is eliminated; accumulative errors are eliminated through forward and reverse filtering on the pre-processed data, and the accuracy of the filtering result is guaranteed; the traffic data flow is accurately forecasted through the single-step forecasting method, and a forecasting result has certain objectivity.

The invention discloses a method for synchronizing time. Single exponential smoothing filtering is respectively carried out on the measured value of a phase position and the measured value of frequency, only last calculation results and the single measured values of this time are used in the calculation process every time, the calculated amount is greatly reduced, and through continuous iterative operations, an algorithm can be stably converged to be close to a statistical expected value. Through the algorithm of the method, low hardware cost can be used, random time errors output by a global positioning system can be compensated in a high-precision mode, the frequency of a high and stable crystal oscillator is measured precisely, and supports are provided for designing a high-precision clock and improving the punctuality accuracy of the clock.