WO2019171106A1 - Prediction-based modelling service in cloud - Google Patents

Prediction-based modelling service in cloud Download PDF

Info

Publication number
WO2019171106A1
WO2019171106A1 PCT/IB2018/051368 IB2018051368W WO2019171106A1 WO 2019171106 A1 WO2019171106 A1 WO 2019171106A1 IB 2018051368 W IB2018051368 W IB 2018051368W WO 2019171106 A1 WO2019171106 A1 WO 2019171106A1
Authority
WO
WIPO (PCT)
Prior art keywords
parameters
applications
here
temporal difference
algorithm
Prior art date
Application number
PCT/IB2018/051368
Other languages
French (fr)
Inventor
Pratik Sharma
Original Assignee
Pratik Sharma
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Pratik Sharma filed Critical Pratik Sharma
Priority to PCT/IB2018/051368 priority Critical patent/WO2019171106A1/en
Publication of WO2019171106A1 publication Critical patent/WO2019171106A1/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound

Definitions

  • a star (A*) algorithm is a best-first search algorithm and its main drawback is its huge memory requirement.
  • the above kind of predictive modelling service can be used for weather predictor applications, automobile insurance applications, health care applications and algorithmic trading applications where predictive models can be built for different assets like stocks, futures, currencies, commodities etc.

Abstract

Here we are provided by the client applications a set of parameters and their valid ranges along with set of values for these parameters from time to time. Here we apply the Temporal Difference learning or the reinforcement learning approach to predict future values of the key parameters of the client application. Also in this invention we try to separate material parameters from non-material ones to make training algorithms much simpler and faster so that they adapt quickly and provide with faster results. We use Minimax Search Algorithm here for the same which searches forward confined to a fixed depth taking into account the past data of each parameter whether it is increasing or decreasing or roughly in a small fixed range or interval.

Description

Prediction-Based Modelling Service In Cloud
In this invention we are provided by the client applications a set of parameters and their valid ranges along with set of values for these parameters from time to time. Here we apply the Temporal Difference learning or the reinforcement learning approach to predict future values of the key parameters of the client application. As a prediction method, Temporal Difference learning considers that subsequent
predictions are often correlated in some sense. The core idea of Temporal Difference learning is that one adjusts predictions to match other, more accurate, predictions about the future. Also in this invention we try to separate material parameters from non-material ones to make training algorithms much simpler and faster so that they adapt quickly and provide with faster results. We use Minimax Search Algorithm here for the same which searches forward confined to a fixed depth (range of each parameter is not violated) taking into account the past data of each parameter whether it is increasing or decreasing or roughly in a small fixed range or interval. We can also use A Star (A*) algorithm which combines features of uniform-cost search and pure heuristic search to efficiently filter all the material parameters of the client application based on past data. A star (A*) algorithm is a best-first search algorithm and its main drawback is its huge memory requirement. The above kind of predictive modelling service can be used for weather predictor applications, automobile insurance applications, health care applications and algorithmic trading applications where predictive models can be built for different assets like stocks, futures, currencies, commodities etc.

Claims

Claims Following is the claim for this invention: -
1. In this invention we are provided by the client applications a set of parameters and their valid ranges along with set of values for these parameters from time to time. Here we apply the Temporal Difference learning or the reinforcement learning approach to predict future values of the key parameters of the client application. As a prediction method, Temporal Difference learning considers that subsequent predictions are often correlated in some sense. The core idea of Temporal Difference learning is that one adjusts predictions to match other, more accurate, predictions about the future. Also in this invention we try to separate material parameters from non-material ones to make training algorithms much simpler and faster so that they adapt quickly and provide with faster results. We use Minimax Search Algorithm here for the same which searches forward confined to a fixed depth (range of each parameter is not violated) taking into account the past data of each parameter whether it is increasing or decreasing or roughly in a small fixed range or interval. We can also use A Star (A*) algorithm which combines features of uniform-cost search and pure heuristic search to efficiently filter all the material parameters of the client application based on past data. A star (A*) algorithm is a best-first search algorithm and its main drawback is its huge memory requirement. The above kind of predictive modelling service can be used for weather predictor applications, automobile insurance
applications, health care applications and algorithmic trading applications where predictive models can be built for different assets like stocks, futures, currencies, commodities etc. The above novel technique by which we provide prediction- based modelling service in cloud for different client applications is the claim for this invention.
PCT/IB2018/051368 2018-03-04 2018-03-04 Prediction-based modelling service in cloud WO2019171106A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/IB2018/051368 WO2019171106A1 (en) 2018-03-04 2018-03-04 Prediction-based modelling service in cloud

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/IB2018/051368 WO2019171106A1 (en) 2018-03-04 2018-03-04 Prediction-based modelling service in cloud

Publications (1)

Publication Number Publication Date
WO2019171106A1 true WO2019171106A1 (en) 2019-09-12

Family

ID=67846484

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2018/051368 WO2019171106A1 (en) 2018-03-04 2018-03-04 Prediction-based modelling service in cloud

Country Status (1)

Country Link
WO (1) WO2019171106A1 (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130110750A1 (en) * 2011-09-28 2013-05-02 Causata Inc. Online temporal difference learning from incomplete customer interaction histories

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130110750A1 (en) * 2011-09-28 2013-05-02 Causata Inc. Online temporal difference learning from incomplete customer interaction histories

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
BAXTER, JONATHAN ET AL.: "TDLeaf(lambda): Combining Temporal Difference Learning with Game-Tree Search", ARXIV.ORG, 5 January 1999 (1999-01-05), XP080649531, Retrieved from the Internet <URL:https://arxiv.org/abs/cs/9901001> [retrieved on 20180705] *

Similar Documents

Publication Publication Date Title
US20220351091A1 (en) Training distilled machine learning models
Budzianowski et al. Sub-domain modelling for dialogue management with hierarchical reinforcement learning
US20160035344A1 (en) Identifying the language of a spoken utterance
Azevedo et al. DRhoGEF2 regulates cellular tension and cell pulsations in the Amnioserosa during Drosophila dorsal closure
CN109697977B (en) Speech recognition method and device
KR101881731B1 (en) Apparatus and Method for Dementia Prediction using Machine Learning, and Storage Medium Storing the Same
Raja et al. Demand based crop recommender system for farmers
US10445653B1 (en) Evaluating reinforcement learning policies
WO2019019346A1 (en) Asset allocation strategy acquisition method and apparatus, computer device, and storage medium
Zafeiriou et al. Short-term trend prediction of foreign exchange rates with a neural-network based ensemble of financial technical indicators
US9929909B2 (en) Identifying marginal-influence maximizing nodes in networks
Handa Prediction of foreign exchange rate using regression techniques
US20170116538A1 (en) System and method for identifying composition preferences
WO2019171106A1 (en) Prediction-based modelling service in cloud
Doumpos Financial decision making using computational intelligence
US20160171608A1 (en) Methods and systems for finding similar funds
Groppo et al. Predicting time-series for water demand in the big data environment using statistical methods, machine learning and the novel analog methodology dynamic time scan forecasting
Daniel Financial time series data processing for machine learning
CN109410005A (en) One kind looking into valence method and looks into valence device
Virk Prediction of bitcoin price using data mining
McDonald et al. A comparison of forecasting approaches for capital markets
Boyer et al. Implications of a reserve price in an agent-based common-value auction
Ayeelyan et al. Prediction strategies of stock market data using deep learning algorithm
Bidyuk et al. Forecasting the volatility of financial processes with conditional variance models
Kapoor et al. Software cost estimation techniques–a review of literature

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18908309

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18908309

Country of ref document: EP

Kind code of ref document: A1