Systems and methods for time series analysis techniques utilizing count data sets

Abstract

Systems and methods are included for adjusting a set of predicted future data points for a time series data set including a receiver for receiving a time series data set. One or more processors and one or more non-transitory computer readable storage mediums containing instructions may be utilized. A count series forecasting engine, utilizing the one or more processors, generates a set of counts corresponding to discrete values of the time series data set. An optimal discrete probability distribution for the set of counts is selected. A set of parameters are generated for the optimal discrete probability distribution. A statistical model is selected to generate a set of predicted future data points. The set of predicted future data points are adjusted using the generated set of parameters for the optimal discrete probability distribution in order to provide greater accuracy with respect to predictions of future data points.

What is claimed is: 
     
         1 . A computer program product, tangibly embodied in a non-transitory machine-readable storage medium, including instructions operable to cause a data processing apparatus to perform operations including:
 receiving a time series data set, wherein the time series data set includes a plurality of data points that correspond to a plurality of discrete values;   generating a set of counts for the time series data set by analyzing the time series data set, wherein a count corresponds to a number of instances of a particular discrete value in the time series data set;   automatically selecting an optimal discrete probability distribution for the set of counts from a set of discrete probability distributions based on a selection criterion;   generating a set of parameters corresponding to the optimal discrete probability distribution;   selecting a statistical model for the time series data set, wherein selecting the statistical model includes using a set of statistical models and the selection criterion;   generating a set of predicted future data points for the time series data set, wherein generating the set of predicted future data points includes using the selected statistical model;   adjusting the set of predicted future data points for the time series data set, wherein adjusting the set of predicted future data points includes using the set of parameters corresponding to the optimal discrete probability distribution; and   predicting a future data point based on the adjusted set of predicted future data points and data input associated with the time series data set.   
     
     
         2 . The computer program product of  claim 1 , wherein a discrete probability distribution outputs a corresponding non-negative value for each value in a range of the discrete probability distribution, and wherein a set of candidate discrete probability distributions includes at least one of a binomial probability distribution, a zero-modified negative binomial distribution, a zero-modified binomial probability distribution, a geometric probability distribution, a zero-modified geometric probability distribution, a Poisson probability distribution, a zero-modified Poisson probability distribution, a log series distribution, or a negative binomial probability distribution. 
     
     
         3 . The computer program product of  claim 1 , wherein the statistical model is configured to utilize the plurality of data points as continuous variables. 
     
     
         4 . The computer program product of  claim 1 , wherein the plurality of data points correspond to a plurality of distinct values, the instructions further comprising providing the predicted future data point to a graphical user interface. 
     
     
         5 . The computer program product of  claim 1 , wherein the time series data set includes greater than a threshold number of data points that correspond to a zero value, and wherein the set of statistical models includes at least one time series model comprising an exponential smoothing model. 
     
     
         6 . The computer program product of  claim 1 , wherein the data processing apparatus comprises an event stream processing engine. 
     
     
         7 . The computer program product of  claim 1 , wherein the data input is received from an edge computing device. 
     
     
         8 . The computer program product of  claim 1 , wherein the data processing apparatus is configured to receive the data input from a sensor device. 
     
     
         9 . The computer program product of  claim 1 , wherein the operations further include:
 receiving the set of discrete probability distributions and the selection criterion from a device, wherein the set of parameters includes at least one of a standard error, an upper limit corresponding to a confidence range, and a lower limit correspond to the confidence range.   
     
     
         10 . The computer program product of  claim 1 , where the data processing apparatus is a component of a home appliance device, an office-related device, a media-playback device, a lighting device, a security system device, a networked energy grid device, or an automobile device. 
     
     
         11 . The computer program product of  claim 1 , wherein selecting the statistical model for the time series data set includes:
 receiving indication of the selection criterion;   generating the set of predicted future data points for the time series data set using a plurality of statistical models; and   generating, by the data processing apparatus, a plurality of error measurements corresponding to each of the plurality of statistical models.   
     
     
         12 . The computer program product of  claim 11 , wherein selecting the statistical model for the time series data set using the plurality of error measurements includes selecting an optimal error measurement of the plurality of error measurements, and wherein the selected statistical model corresponds to the optimal error measurement. 
     
     
         13 . The computer program product of  claim 11 , wherein automatically selecting the optimal discrete probability distribution for the set of counts includes:
 receiving, by the data processing apparatus, indication of selection of the selection criterion;   generating, by the data processing apparatus, a set of candidate discrete probability distributions for the set of counts using a set of discrete probability distribution algorithms; and   generating a plurality of error measurements corresponding to each of the set of candidate discrete probability distributions.   
     
     
         14 . The computer program product of  claim 13 , wherein the received selection criterion includes one of an Akaike information criterion, a weighted least-squares criterion, a log-likelihood criterion, or a Bayesian information criterion. 
     
     
         15 . The computer program product of  claim 13 , wherein automatically selecting the optimal discrete probability distribution for the set of counts using the plurality of error measurements includes selecting an optimal error measurement of the plurality of error measurements, and wherein the optimal discrete probability distribution corresponds to the optimal error measurement. 
     
     
         16 . The computer program product of  claim 1 , wherein the time series data set includes greater than a threshold number of data points that correspond to a zero value, and wherein the set of statistical models includes at least one time series model comprising an exponential smoothing model. 
     
     
         17 . A device comprising:
 a sensor;   a connection component for a connection to a network;   an event stream processor; and   a non-transitory machine-readable storage medium connected to the event stream processor and the sensor, the non-transitory machine-readable storage medium including instructions operable to a data processing apparatus to perform operations including:
 receiving, from the sensor or the connection component, a time series data set, wherein the time series data set includes a plurality of data points that correspond to a plurality of discrete values; 
 generating, with the event stream processor, a set of counts for the time series data set by analyzing the time series data set, wherein a count corresponds to a number of instances of a particular discrete value in the time series data set; 
 automatically selecting, with the event stream processor, an optimal discrete probability distribution for the set of counts from a set of discrete probability distributions based on a selection criterion; 
 generating, with the event stream processor, a set of parameters corresponding to the optimal discrete probability distribution; 
 selecting, with the event stream processor, a statistical model for the time series data set, wherein selecting the statistical model includes using a set of statistical models and the selection criterion; 
 generating, with the event stream processor, a set of predicted future data points for the time series data set, wherein generating the set of predicted future data points includes using the selected statistical model; 
 adjusting, with the event stream processor, the set of predicted future data points for the time series data set, wherein adjusting the set of predicted future data points includes using the set of parameters corresponding to the optimal discrete probability distribution; 
 predicting, with the event stream processor, a future data point based on the adjusted set of predicted future data points and data input associated with the time series data set; and 
 providing, to the sensor or the connection component, the predicted future data point. 
   
     
     
         18 . The device of  claim 17 , wherein the connection component includes an intra-network connection component or an internetwork connection component, and wherein the sensor is operable to sense environmental data. 
     
     
         19 . The device of  claim 17 , wherein selecting the statistical model for the time series data set further includes:
 receiving indication of the selection criterion;   generating the set of predicted future data points for the time series data set using a plurality of statistical models; and   generating, by the data processing apparatus, a plurality of error measurements corresponding to each of the plurality of statistical models.   
     
     
         20 . The device of  claim 18 , wherein selecting the statistical model for the time series data set using the plurality of error measurements includes selecting an optimal error measurement of the plurality of error measurements, and wherein the selected statistical model corresponds to the optimal error measurement. 
     
     
         21 . The device of  claim 19 , wherein automatically selecting the optimal discrete probability distribution for the set of counts includes:
 receiving, by the data processing apparatus, indication of selection of the selection criterion;   generating, by the data processing apparatus, a set of candidate discrete probability distributions for the set of counts using a set of discrete probability distribution algorithms; and   generating a plurality of error measurements corresponding to each of the set of candidate discrete probability distributions.   
     
     
         22 . The device of  claim 19 , wherein:
 the received selection criterion includes one of an Akaike information criterion, a weighted least-squares criterion, a log-likelihood criterion, or a Bayesian information criterion,   automatically selecting the optimal discrete probability distribution for the set of counts using the plurality of error measurements includes selecting an optimal error measurement of the plurality of error measurements, and   the optimal discrete probability distribution corresponds to the optimal error measurement.   
     
     
         23 . A computer-implemented method for adjusting a set of predicted future data points for a time series data set, comprising:
 receiving a time series data set, wherein the time series data set includes a plurality of data points that correspond to a plurality of discrete values;   generating, by a data processing apparatus, a set of counts for the time series data set by analyzing the time series data set, wherein a count corresponds to a number of instances of a particular discrete value in the time series data set;   automatically selecting, by the data processing apparatus, an optimal discrete probability distribution for the set of counts from a set of discrete probability distributions based on a selection criterion;   generating, by the data processing apparatus, a set of parameters corresponding to the optimal discrete probability distribution;   selecting, by the data processing apparatus, a statistical model for the time series data set, wherein selecting the statistical model includes using a set of statistical models and the selection criterion;   generating, by the data processing apparatus, the set of predicted future data points for the time series data set, wherein generating the set of predicted future data points includes using the selected statistical model;   adjusting, by the data processing apparatus, the set of predicted future data points for the time series data set, wherein adjusting the set of predicted future data points includes using the set of parameters corresponding to the optimal discrete probability distribution; and   predicting, by the data processing apparatus, a future data point based on the adjusted set of predicted future data points and data input associated with the time series data set,   wherein the data processing apparatus is operable with a count series forecasting engine in a computing environment.   
     
     
         24 . The computer-implemented method of  claim 23 , wherein a discrete probability distribution outputs a corresponding non-negative value for each value in a range of the discrete probability distribution, and wherein a set of candidate discrete probability distributions includes at least one of a binomial probability distribution, a zero-modified negative binomial distribution, a zero-modified binomial probability distribution, a geometric probability distribution, a zero-modified geometric probability distribution, a Poisson probability distribution, a zero-modified Poisson probability distribution, a log series distribution, or a negative binomial probability distribution. 
     
     
         25 . The computer-implemented method of  claim 23 , wherein the statistical model uses the plurality of data points as continuous variables. 
     
     
         26 . The computer-implemented method of  claim 23 , wherein the plurality of data points correspond to a plurality of distinct values. 
     
     
         27 . The computer-implemented method of  claim 23 , wherein selecting the statistical model for the time series data set includes:
 receiving indication of the selection criterion;   generating the set of predicted future data points for the time series data set using a plurality of statistical models; and   generating, by the data processing apparatus, a plurality of error measurements corresponding to each of the plurality of statistical models.   
     
     
         28 . The computer-implemented method of  claim 27 , wherein selecting the statistical model for the time series data set using the plurality of error measurements includes selecting an optimal error measurement of the plurality of error measurements, and wherein the selected statistical model corresponds to the optimal error measurement. 
     
     
         29 . The computer-implemented method of  claim 27 , wherein automatically selecting the optimal discrete probability distribution for the set of counts further comprises:
 receiving, by the data processing apparatus, indication of selection of the selection criterion;   generating, by the data processing apparatus, a set of candidate discrete probability distributions for the set of counts using a set of discrete probability distribution algorithms; and   generating a plurality of error measurements corresponding to each of the set of candidate discrete probability distributions.   
     
     
         30 . The computer-implemented method of  claim 29 , wherein the received selection criterion includes one of an Akaike information criterion, a weighted least-squares criterion, a log-likelihood criterion, or a Bayesian information criterion, and wherein the data processing apparatus comprises an edge computing device in the computing environment.