Back to results
US 11,893,069 B2Granted
Platform, method, and system for a search engine of time series data
OwnerJencir Leeinventor
Life of this patent
Estimating the term…
Abstract
A system, method, and computer program product provide a search engine which may use time series of data relevant to a search query to generate relevant results that may be used for predictions of events related to the query. Embodiments determine the underlying driving factors that most influenced the search query topic. In one aspect, the predictive ability of time series may be used to determine the driving factors. In another aspect, the system and method may employ a neural network to run a prediction model for determining conditional distribution of future values of the search query topic. The information may be displayed in a user interface where the driving factors and other information may be user edited to observe the effects source data has on the search query topic.
Claims14
A computer program product for an online search engine generating search query results, the computer program product comprising a non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code being configured, when executed by a processor, to:
- generate a user interface (UI) on a computing device with an electronic display;
- generate a search query field configured to receive a user's search query topic;
- receive the search query topic;
- retrieve from one or more online data sources, values of the search query topic, wherein the values of the search query topic are assembled as a primary set of time series data;
- retrieve values of a plurality of factors associated with the search query topic, wherein:
- the plurality of factors are distinct from the search query topic, and
- the values of the plurality of factors are assembled into secondary sets of time series data, distinct from the values of the search query topic stored as the primary set of time series data;
- determine a predictive ability value of two or more of the secondary sets of time series data, for the primary set of time series data, wherein determining the predictive ability value comprises:
- providing the two or more of the secondary sets of time series data to a prediction model,
- predicting, using the prediction model, a first value for a future value of the primary set of time series data, using a selected one of the secondary sets of time series data,
- predicting, using the prediction model, a second value for the future value of the primary set of time series data, without using the selected one of the secondary sets of time series data, and
- determining a difference between the first value for the future value of the primary set of time series data and the second value for the future value of the primary set of time series data, wherein the difference is the predictive ability value of the selected one of the secondary sets of time series data on the primary set of time series data of the search query topic;
- set a threshold value for the determined predictive ability value;
- rank the secondary sets of time series data based on the determined predictive ability value of respective secondary sets of time series data;
- determine additional predicted future values of time series data for the search query topic based on secondary sets of time series data ranked above the threshold value;
- determine one or more factors with the determined predictive ability value above the threshold value as most influencing the search query topic;
- and
- display the additional future predicted values of time series data for the search query topic and the one or more factors most influencing the search query topic, in the user interface.
15 more elements
A method of generating search query results and operating an online search engine, comprising:
- generating a user interface (UI) on a computing device with an electronic display;
- generating a search query field configured to receive a user's search query topic;
- receiving the search query topic;
- retrieving from one or more online data sources, values of the search query topic, wherein the values of the search query topic are assembled as a primary set of time series data;
- retrieving values of a plurality of factors associated with the search query topic, wherein:
- the plurality of factors are distinct from the search query topic, and
- the values of the plurality of factors are assembled into secondary sets of time series data, distinct from the values of the search query topic stored as the primary set of time series data;
- determining a predictive ability value of two or more of the secondary sets of time series data, for the primary set of time series data, wherein determining the predictive ability value comprises:
- providing the two or more of the secondary sets of time series data to a prediction model,
- predicting, using the prediction model, a first value for a future value of the primary set of time series data, using a selected one of the secondary sets of time series data,
- predicting, using the prediction model, a second value for the future value of the primary set of time series data, without using the selected one of the secondary sets of time series data, and
- determining a difference between the first value for the future value of the primary set of time series data and the second value for the future value of the primary set of time series data, wherein the difference is the predictive ability value of the selected one of the secondary sets of time series data on the primary set of time series data of the search query topic;
- setting a threshold value for the determined predictive ability value;
- rank the secondary sets of time series data based on the determined predictive ability value of respective secondary sets of time series data;
- determining additional predicted future values of time series data for the search query topic based on secondary sets of time series data ranked above the threshold value;
- determining one or more factors with the determined predictive ability value above the threshold value as most influencing the search query topic;
- and
- displaying the additional future predicted values of time series data for the search query topic and the one or more factors most influencing the search query topic, in the user interface.
15 more elements
An online host server, comprising:
- a network connection;
- a memory storage device;
- and a processor coupled to the network connection and connected to the memory storage device, wherein computer executable instructions in the memory storage device configure the processor to:
- generate a user interface (UI) on a computing device with an electronic display;
- generate a search query field configured to receive a user's search query topic;
- receive the search query topic;
- retrieve from one or more online data sources, values of the search query topic, wherein the values of the search query topic are assembled as a primary set of time series data;
- retrieve values of a plurality of candidate factors associated with the search query topic, wherein:
- the plurality of candidate factors are distinct from the search query topic, and they are obtained by:
- assembling the values of the plurality of candidate factors into random secondary sets of time series data;
- determining an optimized quantile regression loss of two or more of random secondary sets of time series data, for the primary set of time series data wherein for each secondary random set, a quantile regression is run resulting in a single number;
- and computing a predictive ability value for every candidate factor from the optimized quantile regression losses;
- set a threshold for the determined predictive ability value;
- rank the candidate factors by the determined predictive ability value;
- determine one or more candidate factors with the determined predictive ability value above the threshold as most influencing the search query topic;
- determine a predicted conditional distribution of a future value of the search query topic at a forecast horizon value of time series data for the search query topic, comprising the candidate factors with the predicative ability value ranked above the threshold;
- and Filing Date:
- Mar. 9, 2021 display the predicted conditional distribution of the future value at the forecast horizon of time series data for the search query topic and the one or more candidate factors most influencing the search query topic, in the user interface.
15 more elements
Dependencies are inferred from the claim text.