多元预测：通往更准确预测之路 [THR3637].pdf

1、 Multivariate Predictions:Your Pathway to More Accurate ForecastsTHR3637Mark RinaldiSenior DirectorEPM Product ManagementOctober 14,2025The Forecasting ChallengeManual Forecasting Relies Heavily on Human Intuition(Heuristics)Historical Trends Alone Miss External InfluencesUnivariate=Single Input-Lim

2、ited ViewMultivariate=Multiple Correlated Inputs-Higher AccuracyExample:Sales+Price+Promotions=Richer Prediction2Copyright 2025,Oracle and/or its affiliatesRicher Sales PredictionPromotionsSalesPriceActualsForecast123456789101112123456789101112Univariate PredictionProduct Sales VolumesActualsForecas

3、t123456789101112123456789101112Multivariate PredictionProduct Sales VolumesHistorical dataPredictionHistorical dataPredictionContextual Factors:-Mktg Promotions-Average Sales Price-EtcOperational Drivers:-Industry Volumes-GDP RateWhat-if Projections(for Investment Decision-making)3Copyright 2025,Ora

4、cle and/or its affiliatesUse Advanced Predictions for Investment Decision MakingMultivariate PredictionsEnables More Powerful PredictionsCorrelates with Other Provided Data PointsLeverages More Sophisticated AlgorithmsImproves AccuracyEasy to Configure&UseStep-by-Step Configuration WizardAutomated S

5、election of AlgorithmsDesigned for Finance Users4Copyright 2025,Oracle and/or its affiliatesConfiguration WizardNew Prediction TypeExplainabilityExplainability:Feature ImportanceBuild trust in the AI with Explainability5Copyright 2025,Oracle and/or its affiliatesThe relative impact each variable has

6、 on the predictionIntelligent Variable SelectionOptimizing the right variables6Copyright 2025,Oracle and/or its affiliatesToo many low-value variables can degrade the quality of a prediction“Feature Selection”is a method of choosing the most relevant features(variables)from a dat