KR20230052010A - Demand forecasting method using ai-based model selector algorithm - Google Patents

Abstract

The present invention discloses a demand prediction method. More specifically, the demand prediction method using an AI-based model selection algorithm which uses big data to generate and operate an advanced prediction model which predicts the demand for parts, materials, or the like required by production companies in various fields. According to an embodiment of the present invention, multiple prediction scenarios are derived by using multiple data sources. By performing learning and evaluation procedures using multiple statistical and machine learning techniques and presenting an ensemble model which presents the optimal predicted demand, it has the effect of providing optimal prediction results for the current situation of the company.

Description

Demand forecasting method using AI-based model selection algorithm {DEMAND FORECASTING METHOD USING AI-BASED MODEL SELECTOR ALGORITHM}

The present invention relates to a demand forecasting method, and in particular, to a demand forecasting method using an AI-based model selection algorithm that predicts the demand for service parts and finished products in various industrial fields using AI-based big data analysis methodologies and optimizes them again. it's about

In a company that produces and manufactures various products, it is very important to predict the demand for parts for production and the customer's future demand for the finished product to be produced.

This is because if the predicted demand quantity is inaccurate, production plans may be disrupted, and problems such as bad inventory due to excessive production may increase economic or time waste.

For example, air-conditioning and heating system-related parts vary greatly in demand depending on the season. That is, it can be easily predicted that the demand for cooling devices such as air conditioners and electric fans will greatly increase in summer, and the demand for heating devices such as warmers and boilers will greatly increase in winter.

However, various variables exist in this simple prediction. Even in summer, when the average temperature is lower than normal, there is a possibility that the actual demand will be significantly lower than the predicted demand. In winter, when the average temperature is significantly lower than normal or there are frequent heavy snowfalls, there is a possibility that the actual demand will greatly exceed the predicted demand. exist.

Therefore, if the amount of demand for air conditioning and heating devices and air conditioning related parts required for each season is accurately predicted, companies will be able to greatly reduce cost waste due to excessive or insufficient production of air conditioning devices.

Accordingly, various techniques for predicting future demand for specific parts and products have been proposed in the fields of production and manufacturing. However, the demand forecasting techniques disclosed so far have been generally used to predict the future demand by using only the past demand with respect to a specific part or product for which the future demand is to be predicted. In the case of predicting future demand through this technique, there is a limitation in that the accuracy of the prediction is lowered if the past data of the specific part or object to be predicted is not sufficient along with the problem of not fully utilizing the given information.

In addition, various demand forecasting methodologies based on machine learning and deep learning are presented in addition to statistical techniques that use only past data when forecasting demand, but these AI-based models are also based on a single model, so there are limitations in predicting multiple items exists.

In addition, prediction methods using multiple models rather than a single AI model currently presented also have difficulty in selecting an appropriate model due to issues such as overfitting according to the model selection method of the fitting criterion.

Registered Patent Publication No. 10-1966557 (Announcement date: 2019.04.05.)

The present invention has been made to overcome the above-mentioned limitations, and the present invention presents various forecast scenarios and records past actual forecast results in predicting future demand for parts or products in production and manufacturing companies. There is a task to provide a demand forecasting method using an AI-based model selection algorithm that can derive optimal results by presenting an ensemble algorithm that derives optimal prediction results based on further consideration.

In order to solve the above problems, the demand forecasting method using an AI-based model selection algorithm in an embodiment of the present invention queries the demand forecast result for one or more parts operated by the company from the company terminal belonging to a specific industry field. Receiving step, extracting and processing a plurality of data sources for demand performance information, meta information and predictive factors related to any one part of the company or a segment consisting of a combination of a plurality of parts from the data warehouse, and processing step for each part Generating a plurality of forecast scenarios including prediction models such as one or more statistical techniques and machine learning techniques based on a plurality of data sources, and storing demand forecast results according to a plurality of forecast scenarios through a model selection unit in the past actual forecast result history In addition, generating an optimal ensemble model at the current forecasting point by evaluating a plurality of forecast scenarios and reflecting the forecast error level, and deriving the final demand through the ensemble model, and modifying the result as needed by the user. steps may be included.

The prediction scenario may include a level for a learning unit, a demand prediction algorithm and model, and learning factor values input to the prediction algorithm and model.

The predictive model may be based on one or more of a plurality of time series models, data mining models, machine learning models, and deep learning models.

After generating the plurality of prediction scenarios, detecting outliers out of the normal range for the prediction result, and removing the quantity of data promotion, designating a promotion flag, or smoothing the data promotion according to the detection result and a pre-calculated error. Alternatively, the method may further include correcting the prediction result through a preprocessing procedure including upper and lower limits.

In the step of deriving the final demand forecast result through the ensemble model, one ensemble model is generated by evaluating each of the derived demand forecast results, and the demand forecast result by analyzing the plurality of forecast scenarios is displayed through a user interface (UI). The method may further include providing the information to the enterprise terminal through and modifying a prediction result or generating and adding a prediction scenario by a person in charge of the enterprise through the user interface.

The generating of the ensemble model may include deriving an ensemble result through evaluation for each scenario using an AI algorithm or a machine learning algorithm.

According to an embodiment of the present invention, an ensemble model that derives a plurality of prediction scenarios by utilizing a plurality of data sources, performs learning and evaluation procedures using a plurality of statistical techniques and machine learning techniques, and presents an optimal predicted demand amount. By deriving, there is an effect of providing prediction results appropriate to the current situation of the company.

In addition, according to an embodiment of the present invention, in addition to various data sources used for analysis, past actual prediction result history is further utilized to overcome the overfitting issue of the algorithm and to derive and provide higher-dimensional demand forecast results. It works.

1 is a diagram illustrating a demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention.
2 is a table showing an example of data sources used in a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.
3 is a table showing an example of prediction scenarios used in a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.
4 is a diagram showing the structure of a system implementing a demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention.
5 is a diagram illustrating an AI-based model selection algorithm according to an embodiment of the present invention.
6 is a diagram for explaining a conventional outlier preprocessing procedure.
7 is a diagram showing a prediction result correction algorithm of a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.
8 is a diagram for explaining a user simulation function provided by a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.

In the entire specification, when a part is said to "comprise" or "include" a certain component, it means that it may further include other components, not excluding other components unless otherwise stated. In addition, terms such as "...Unit" and "...System" described in the specification mean a unit that processes one or more functions or operations, which are hardware, software, or hardware And it can be implemented as a combination of software.

In addition, the term "embodiment" in this specification means to serve as an illustration, example, or illustration, but the subject matter of the invention is not limited by such examples. Also, while "comprising", "comprising", "having" and other similar terms are used, when used in the claims, the transition word "comprising" is an open term that does not exclude any additional or different elements. It can be used generically in a similar way to the term "comprising".

The various techniques described herein may be implemented with hardware or software, or a combination of both where appropriate. As used herein, the terms "...Unit" and "...System" likewise refer to a computer-related entity, i.e., hardware, a combination of hardware and software, software, or execution. It can be treated as equivalent to the city's software. In addition, each function executed in the program of the present invention may be configured in module units and recorded in one physical memory or distributed between two or more memories and recording media.

In the following description, the 'demand forecasting method using an AI-based model selection algorithm' according to an embodiment of the present invention may be referred to as a 'demand forecasting method' for convenience of description, and may be referred to as a 'demand forecasting method using an AI-based model selection algorithm' The system' may be expressed as a 'demand forecasting system' or a 'system'.

Hereinafter, a demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention will be described with reference to the drawings.

1 is a diagram showing a demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention, and FIG. 2 is a table showing an example of data sources used in a demand forecasting system using a model selection unit according to an embodiment of the present invention. 3 is a diagram showing an example of a forecast scenario used in a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.

In each of the following steps, the execution subject is the demand forecasting system according to the embodiment of the present invention and each component constituting the same, even if there is no separate description, and each step can be implemented in the form of a computer program using a known programming language. and can be recorded on a recording medium capable of reading and writing.

First, referring to FIG. 1, in the demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention, a demand prediction result for one or more parts operated by the company is inquired from a company terminal of a company belonging to a specific industry field. Receiving step (S100), extracting and processing a plurality of data sources for demand performance information, meta information, predictors, etc. related to any one part of the company or a segment consisting of a combination of a plurality of parts from the data warehouse ( S110), generating a plurality of prediction scenarios including prediction models such as one or more statistical techniques and machine learning techniques based on a plurality of data sources for each part (S120), A step of generating an optimal ensemble model at the current forecasting point by evaluating a plurality of forecast scenarios based on the demand forecast result based on the history of past actual forecast results and reflecting the forecast error level (S130), and the final demand quantity through the ensemble model It may include a step (S140) of deriving and allowing the user to modify the result as needed.

In the step of receiving a query from a company terminal for a demand prediction result for one or more parts operated by the company (S100), as a company belonging to a specific industry introduces the system of the present invention, a person in charge of the company uses his/her terminal to use the system This is the step of accessing and querying the demand forecast results for one or more parts operated by the company.

The demand forecasting system according to an embodiment of the present invention is a solution for companies that want SCM consulting such as big data-based demand forecasting, sales planning, sales management, and inventory optimization in each industry field, and companies that have introduced the system of the present invention The person in charge of the company can query the demand forecast for parts operated by the company.

Next, the step of extracting and processing a plurality of data sources related to a part or a segment composed of a combination of a plurality of parts from the data warehouse ( S110 ) is a step in which the system acquires the part-related data source. In this step, the system extracts data sources stored in the data warehouse to generate forecast scenarios, especially various data sources related to the parts of the company to be analyzed, so that all available data related to the parts of the company can be utilized. do. Such data sources include demand performance information, meta information, predictive factors, and the like, as illustrated in FIG. 2 . Demand performance information (a) can be data on monthly parts demand and can be used as base data for time series forecasting. In addition, the meta information (b) may include master part attribute data, and may be used as a segment analysis and prediction factor by two or more part data. Here, a segment refers to a combination of two or more parts for analysis.

In addition, the system according to another embodiment of the present invention may further utilize data such as an external data source (c) as a data source in addition to the above-described data. Such external data sources include time events, weather information, crawling data, and data provided by companies in the same or similar fields, and a person in charge of the company may manually input desired data through the company terminal.

In particular, the system according to the embodiment of the present invention can further utilize the prediction history (d) derived by the past prediction algorithm in addition to the above-mentioned data sources for analysis, and use it for correction of prediction results and model selection algorithm. can be utilized

Next, in the step of generating a plurality of prediction scenarios by performing statistical techniques and machine learning techniques based on a plurality of data sources (S120), a plurality of prediction scenarios are generated through various algorithms and user simulations using the various information listed in step S110. It is a step of deriving a prediction scenario of

It is possible to detect abnormal data and analyze data patterns such as trends, seasonality, and intermittent data through the above-mentioned demand performance information, meta information, and predictors for parts or segments, and the system predicts parts or segments from these data sources. You can create scenarios.

The prediction scenarios above may include buckets, and may be composed of levels, models, and model parameters. A bucket means a data cycle such as year, month, or week to be predicted, and a prediction model ( Alternatively, a prediction result may be derived by deriving a prediction module) and determining and inputting model parameters to be input to the prediction model.

The data source may have various patterns, and the system may derive prediction results for parts or segments by applying either a statistical technique or a machine learning technique suitable for the data pattern for each scenario. Alternatively, a prediction result may be generated by ensembling a plurality of prediction results.

Statistical techniques include time-series analysis and data mining analysis, and machine learning techniques include general machine learning and deep learning algorithms such as neural networks.

Known forecasting models for double, time series analysis include Weighted Moving Average, Holt-winter's Exponential Moving Average, Croston Model, ARIMA, and Seanonal Native.

In addition, predictive models for data mining analysis include linear models, generalized linear models, and decision trees.

In addition, prediction models for general machine learning analysis include boosting models such as gbm, xgboost, light gbm, and catboost, and bagging models such as randomForest and decision forest.

In addition, as a predictive model for deep learning analysis, along with ANN (Artificial Neural Network) and DNN (Deep Neural Network), RNN (Recurrent Neural Network) and LSTM (Long Short-Term Memory) and Long-and Short-Term Temporal Patterns (LSTNet).

The system selects one part or two or more parts segments, and determines each prediction model and model parameters to be input thereto through analysis of related data sources to generate a plurality of prediction scenarios.

In particular, according to an embodiment of the present invention, the system considers dimensional properties such as longitudinal properties of data (Time-series Data), cross-section properties (Cross-section Data), and inter-item reciprocity (Inter-item). At least 10 prediction scenarios are configured to predict the parts of Referring to FIG. 3 , prediction results according to prediction scenarios according to an embodiment of the present invention are illustrated, and prediction results according to each scenario are classified according to prediction models, and monthly data can be checked as numerical values or graphs.

Next, through the model selection unit, the demand forecast results according to the plurality of forecast scenarios are evaluated based on the history of actual forecast results in the past, and a plurality of forecast scenarios are evaluated, and the level of forecast error is reflected to generate an optimal ensemble model at the current forecast time. In step S130, a demand prediction result for a part or segment is derived using the 10 or more prediction scenarios derived in step S120, and the demand prediction result derived through the model selection unit included in the system of the present invention is evaluated. This creates an ensemble model.

According to the system of the present invention, it is characterized by providing a model selection unit for performing an ensemble for a plurality of prediction scenarios. The model selection unit evaluates the forecast results by forecast scenarios and performs ensemble, and reflects the actual forecast error level of the past planning month for each forecast scenario to the analysis to determine the best predictive results can be derived.

In particular, the model selection unit of the present invention selects an optimal model (Best-pick) to solve the overfitting problem caused by selecting a model using only the best-fit level of the past interval within the training data by the conventional analysis system. ), a new concept was applied.

In detail, in order to build an optimal ensemble model, the model selection unit proceeds with an indicator function to determine whether the prediction results for each past prediction scenario are good or bad. It implements the learning mart, which is the learning data for the technique selection algorithm, and integrates the prediction results by conducting scenario evaluation on the current prediction scenario using machine learning and deep learning algorithms.

In particular, when integrating prediction results, the model selector processes an ensemble for prediction scenarios of a certain level or higher probabilistically using a cut-off indicator function.

In addition, the ensemble algorithm of the above-described model selection unit performs machine learning and deep learning processes in units of items and parts segments, and as it is possible to determine different optimal ensemble models for each item and part segment, the accuracy of demand forecasting is improved. .

In addition, according to an embodiment of the present invention, the step of generating an ensemble model in step S130 described above includes the step of detecting an outlier outside the normal range with respect to the prediction result, and the data according to the detection result and the pre-calculated error The method may further include correcting the predicted result through a preprocessing procedure including removing a quantity of promotions, designating a promotion flag, smoothing, or limiting upper and lower limits.

A detailed description of an algorithm for detecting such outliers will be described later.

And, in the step of deriving the final demand forecast result through the generated ensemble model (S140), the analysis result of the demand forecast for the parts operated by the company, etc. can provide

In addition, step S140 is a step of providing a demand forecast result by analyzing a plurality of forecast scenarios to a company terminal through a user interface (UI), and a company manager creating and adding a user scenario for simulation through the user interface. Further steps may be included, and later, the person in charge of the company can check not only the optimal result provided through the system, but also the simulation result according to his or her intention.

At this time, the generated ensemble model is a model selected according to one or more statistical techniques and machine learning techniques in a plurality of prediction scenarios, and more accurate as the ensemble is performed by classifying meaningful scenarios at the current time by evaluating and learning scenarios in the past planning section. Demand forecasting results can be provided.

Hereinafter, a demand forecasting system using an AI-based optimal pick selection algorithm for implementing the above-described demand forecasting method of the present invention will be described in detail with reference to the drawings.

4 is a diagram showing the structure of a system implementing a demand forecasting method using an AI-based model selection algorithm according to an embodiment of the present invention.

Referring to FIG. 4, the demand forecasting system 100 using the AI-based model selection algorithm according to an embodiment of the present invention includes demand performance information 111, meta information 112, and predictor factors 113 related to parts of a company. ) When a demand prediction result for one or more parts operated by the company is queried from the data warehouse 110 storing a plurality of data sources including A plurality of data sources related to any one part of the company or a segment consisting of a combination of parts is extracted, and based on the plurality of data sources for each part, one or more statistical techniques and machine learning techniques are performed to create a predictive model. A scenario derivation unit 120 for generating a plurality of prediction scenarios including a scenario derivation unit 120 for the plurality of prediction scenarios, an outlier detection unit for post-predicting the relationship between an error level and an outlier value using an AI model by statistical techniques and machine learning techniques (130), a model selection unit (140) generating an ensemble model at the current forecasting point by reflecting the forecasting error level of the past planning month based on the past forecasting result history while deriving demand forecasting results according to a plurality of forecasting scenarios. And, it may include an optimum prediction unit 150 for deriving a final demand quantity prediction result through an ensemble model.

The system 100 according to an embodiment of the present invention may be implemented on a base platform including an operating system on which each of the above-described components is executed. Such a base platform may include a framework, a big data platform, a Windows server, and programming languages such as JAVA, C#.net, Python, and R, and data storage means such as a database and storage.

In addition, the service of the present invention can be provided on a web basis, and thus the enterprise terminal accesses the system 100 through the client provided by the demand forecasting system 100 according to the embodiment of the present invention and parts for the enterprise Demand forecast analysis results are available.

The data warehouse 110 may store various types of information for analyzing demand for parts in the system 100 according to an embodiment of the present invention.

In this data warehouse 110, not only the corresponding company, but also information obtained from various companies and external information collected through crawling may be stored, and demand performance information 111 for each part, parts Data sources such as meta information 112 and predictor 113 for may be stored.

In particular, the data warehouse 110 stores past prediction scenarios generated at the time of request from other companies through the system 100 at a previous point in time, and the model selection unit 140 described later can further utilize them when evaluating scenarios.

The scenario derivation unit 120 may generate a plurality of prediction scenarios 121 using various data sources stored in the data warehouse 110 . The prediction scenario 121 may include a bucket, level, model, and model parameters, and the scenario derivation unit 120 selects one part or two or more part segments, and analyzes data sources related thereto to obtain information about each model and model. A plurality of prediction scenarios are generated by determining model parameters to be input thereto.

Here, the reason why the predictive model is configured in various ways is that an optimized model exists according to the characteristics of data to be predicted. As an example, time series analysis such as ARIMA is suitable when past demand performance has only simple volatility, machine learning analysis for event-sensitive data, and Croston analysis for intermittent parts.

In addition, according to an embodiment of the present invention, the outlier detection unit 130 detects an outlier outside the normal range for the prediction result by the prediction scenario 121 and performs preprocessing according to the detection result and a pre-calculated error. may further include.

Anomaly detection is to detect anomalies that may exist in the prediction result derived by the prediction scenario 121, and the anomaly detection unit 130 may be performed by statistical techniques and machine learning techniques, and the error level and Relationships with outliers can be post-corrected using an AI model.

The above-described preprocessing procedure is a procedure for finding and removing points having different values different from the general situation, and changing the intended part demand amount in the data, for example, in the case of a promotion for parts of a company, the promotion quantity It can be removed or a promotion flag can be designated, and if the change is caused by an unintended temporary external factor, it can be processed through data smoothing or upper/lower limit restrictions.

The model selector 140 may determine an ensemble model by evaluating each prediction scenario 121 through an AI-based method selector algorithm, which is an applied technique.

The model selection unit 140 may evaluate the current prediction scenario 121 based on the learning result through the technique selection algorithm in order to build an ensemble model for optimal selection. At this time, history by past prediction scenarios may be evaluated, and an ensemble model may be generated by integrating the prediction results by performing a machine learning technique based on the characteristics of input variables of the prediction scenario.

The final prediction unit 150 may provide a final demand quantity prediction result for a part or segment through the generated ensemble model. The final demand forecast result may include an integrated forecast result and a forecast result for each scenario, and may further provide simulation results according to manual input by a person in charge of the company.

According to the above-described structure, the demand forecasting system according to an embodiment of the present invention derives a plurality of forecast scenarios based on data sources of various viewpoints such as abnormal data and data patterns, and evaluates and evaluates a plurality of forecast models based thereon. By ensembling, it is possible to provide information optimized for each company related to demand forecasting and inventory of parts operated by the company. Demand forecast information can be provided.

Hereinafter, a technique selection algorithm according to an embodiment of the present invention will be described with reference to the drawings.

5 is a diagram showing an optimal AI-based model selection algorithm according to an embodiment of the present invention.

Referring to FIG. 5 , a best-pick based method selector algorithm applied to a model selector according to an embodiment of the present invention is shown. In the present invention, data can be classified on an index-based basis.

First, a plurality of prediction scenarios are extracted from the data warehouse 110 and input mart 203 as learning data is implemented. In particular, according to the technique selection algorithm according to an embodiment of the present invention, among the data sources stored in the data house 110, the actual prediction error level according to the forecasting history 201, which is the actual forecasting scenario of the past planning month, is reflected. To do this, through the indicator function 202, data whose evaluation metrics are greater than or equal to a threshold may be selected and reflected in the input mart 203. This evaluation may be expressed as True/False, Good/Bad, and the like.

Next, the input mart 203 is split for each scenario, input to the machine learning technique or deep learning algorithm 205 as training data 204 to learn the prediction model, and the prediction result by each prediction scenario, that is, , obtain test data 206 for evaluation.

Next, an evaluation score 207 is calculated by evaluating the test data 206, and an ensemble for scenarios of a certain level is processed probabilistically through a cut-off indicator. do.

Hereinafter, an outlier preprocessing algorithm according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

6 is a diagram for explaining a conventional outlier preprocessing procedure, and FIG. 7 is a diagram showing a prediction result correction algorithm of a demand forecasting system using an AI-based model selection algorithm according to an embodiment of the present invention.

Referring to FIG. 6, the outlier in the conventional demand data is in the form of a significantly higher or lower value in terms of demand in a specific period '201805' compared to other periods, as shown in the data as shown in FIG. can appear as

The cause of such an outlier may be a change in demand due to a regular promotion for the corresponding part, or a temporary external factor such as a change in the supply chain or weather, and conventionally, preprocessing is performed in different ways.

In detail, if a regular promotion is the cause, the relevant promotion quantity is removed or a promotion flag is designated to be reflected in the prediction result, and if a temporary external factor is the cause, data smoothing is performed to reduce the overall data Flatten the curvature or set upper and lower limits to remove data outside a specific range.

According to this conventional method, it is difficult to simulate the effect of the actual preprocessing effect on the prediction, and it is difficult to determine whether or not there is an outlier in the input data other than performance.

In order to overcome these limitations, according to the outlier detection algorithm of the demand forecasting system according to an embodiment of the present invention, an error is calculated (301) for the forecast result (300), and the forecast scenario input mart (303) is used as a statistical technique. and an error correction input mart (305), which is learning data for detecting an effect of an outlier on a prediction by inputting the input to the machine learning technique 304 to perform learning to detect an outlier and derive a relationship between the calculated error value and the outlier. ) to create

And, dividing the error correction input mart 305 for each prediction scenario, inputting it to a statistical technique or machine learning technique 307 as training data 306 to train a prediction model, and evaluating the effect of outliers on each prediction scenario Obtain test data 308 for

Then, a corrected prediction result 309 is derived by post-prediction of the prediction result using the evaluation result, and a subsequent optimum pick selection procedure is performed.

Hereinafter, other functions provided by the demand forecasting system using a model selector according to an embodiment of the present invention will be described with reference to the drawings.

8 is a diagram for explaining a user simulation function provided by a demand forecasting system using a model selector according to an embodiment of the present invention.

Referring to FIG. 8, in the demand forecasting system using the AI-based model selection algorithm according to an embodiment of the present invention, in addition to the regular forecasting results provided according to the request of the company subscribed to the system, the company manager directly predicts scenarios when needed by the company. can be created and registered, and analysis can be performed and predicted results can be provided based on the scenario.

The person in charge of the company accesses the system using the company terminal, registers other scenarios by entering a bucket and a plan, etc., as illustrated in the drawing, adds a target part (Item) for each attribute, You can request analysis from the system by adding a predictive model for time series analysis and machine learning analysis, setting parameters to be input to the predictive model, and inputting them to the system.

Accordingly, the system may perform analysis on other scenarios to derive demand forecasting results and provide them to enterprise terminals.

Although many details are specifically described in the above description, this should be interpreted as an example of a preferred embodiment rather than limiting the scope of the invention. Therefore, the invention should not be defined according to the described examples, but should be defined according to the scope of the claims and the scope of the claims.

100: demand forecasting system 110: data warehouse
111: demand performance information 112: meta information
113: predictor 114: prediction history
120: scenario derivation unit 121: prediction scenario
130: outlier detection unit 140: model selection unit
150: final prediction unit

Claims (6)

Receiving a query for a demand prediction result for one or more parts operated by the company from a company terminal of a company belonging to a specific industry field;
Extracting and processing a plurality of data sources including demand performance information, meta information, and predictive factors related to any one part of the company or a segment composed of a combination of a plurality of parts from the data warehouse;
Generating a plurality of predictive scenarios including predictive models by performing one or more statistical techniques and machine learning techniques based on a plurality of data sources for each part;
Generating an ensemble model at a current forecasting point by deriving a demand forecast result according to a plurality of forecasting scenarios through a model selection unit, but reflecting a forecasting error level of a past planning month based on a history of past forecasting results; and
Deriving a final demand forecast result through the ensemble model
A demand forecasting method using an AI-based maximum model selection algorithm that includes. According to claim 1,
The predicted scenario is
A demand forecasting method using an AI-based model selection algorithm that includes a level for a learning unit, a demand forecasting algorithm and model, and a learning factor value input to the forecasting algorithm and model. According to claim 2,
The predictive model,
A method for forecasting demand using an AI-based model selection algorithm based on one or more of a plurality of time series models, data mining models, machine learning models, and deep learning models. According to claim 3,
After generating the plurality of prediction scenarios,
detecting outliers out of the normal range for the prediction result; and
According to the detection result and the pre-calculated error, removing the quantity of data promotion, designating a promotion flag, or correcting the predicted result through a pre-processing procedure including smoothing or upper and lower limits.
Demand forecasting method using an AI-based model selection algorithm further comprising a. According to claim 1,
The step of deriving the final demand forecast result through the ensemble model,
providing a demand forecast result obtained by analyzing the plurality of forecast scenarios to the enterprise terminal through a user interface (UI); and
Step of the enterprise manager creating and adding a prediction scenario through the user interface
Demand forecasting method using an AI-based model selection algorithm further comprising a. According to claim 1,
The step of generating the ensemble model,
Step of deriving ensemble results through scenario-by-scenario evaluation using AI algorithm or machine learning algorithm
A demand forecasting method using an AI-based model selection algorithm that includes.

Images