KR20240025760A - System for forecasting agricultural demand - Google Patents

Abstract

The agricultural product demand prediction system according to an embodiment of the present invention includes a data collection unit that collects relevant data necessary for predicting demand for the agricultural product distribution company from agricultural product distribution companies through a predetermined management server, and the data collected by the data collection unit. A database unit that stores data containing at least one of the item, quantity, and margin of data by item, and analysis that analyzes the item, quantity, and margin of agricultural products based on the data by item stored in the database unit. An item forecasting unit that predicts the higher level items above a preset value based on the expected demand items analyzed in the analysis unit according to a predetermined calculation method, and an item forecasting unit that predicts the expected demand quantity of the item predicted in the item forecasting unit according to a predetermined calculation method. A quantity prediction unit that predicts the quantity within a preset range, and a margin prediction unit that assumes sales of the quantity of the item predicted by the quantity prediction unit, and predicts the seller's sales margin within a preset range according to a preset calculation method. wealth; and a proposal unit that finally calculates the margin of the margin prediction unit for the item predicted in the item prediction unit and the quantity predicted in the quantity prediction unit and proposes a sales quantity to the seller. will be.

Description

Agricultural demand forecasting system {System for forecasting agricultural demand}

The present invention relates to a demand prediction system for agricultural products, and more specifically, to a demand prediction system for agricultural products that improves the efficiency of agricultural product sales by predicting future demand for agricultural products based on agricultural product sales data.

Generally, agricultural products produced at farms go through a complex distribution process before they are provided to consumers.

This conventional agricultural product distribution process had problems in that prices rose to increase the sales margin of agricultural products and the sales process was complicated.

In addition, due to the nature of stores securing large quantities of agricultural products in advance and then selling them to consumers, failure to accurately predict demand could result in significant business losses.

The present invention has designed a demand forecasting system that can reduce sellers' operating losses by predicting more accurate demand by analyzing consumers' past sales items and quantities.

The purpose of the present invention is to improve the efficiency of selling agricultural products through more precise prediction of demand for agricultural products.

The agricultural product demand prediction system according to an embodiment of the present invention includes a data collection unit that collects relevant data necessary for predicting demand for the agricultural product distribution company from agricultural product distribution companies through a predetermined management server, and the data collected by the data collection unit. A database unit that stores data containing at least one of the item, quantity, and margin of data by item, and analysis that analyzes the item, quantity, and margin of agricultural products based on the data by item stored in the database unit. An item forecasting unit that predicts the top items above a preset value based on the expected demand items analyzed in the analysis unit according to a predetermined calculation method, and an item forecasting unit that predicts the expected demand quantity of the item predicted in the item forecasting unit according to a predetermined calculation method. A quantity prediction unit that predicts the quantity within a preset range, and a margin prediction unit that assumes sales of the quantity of the item predicted by the quantity prediction unit, and predicts the seller's sales margin within a preset range according to a preset calculation method. It may include a proposal unit that finally calculates the margin of the margin prediction unit for the item predicted in the item prediction unit and the quantity predicted in the quantity prediction unit and proposes a sales quantity to the seller.

According to the present invention, there is an effect of improving the efficiency of agricultural product sales through more precise prediction of demand for agricultural products.

1 is a block diagram showing the overall configuration of an agricultural product demand prediction system according to an embodiment of the present invention.
Figure 2 is a configuration diagram showing an agricultural product demand prediction system according to an embodiment of the present invention.
Figure 3 is a configuration diagram showing the sales structure of an agricultural product demand prediction system according to an embodiment of the present invention.
4 to 17 are diagrams illustrating a demand prediction model of an agricultural product demand prediction system according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings. However, the spirit of the present invention is not limited to the presented embodiments, and those skilled in the art who understand the spirit of the present invention may add, change, or delete other components within the scope of the same spirit, or create other degenerative inventions or this invention. Other embodiments that are included within the scope of the invention can be easily proposed, but this will also be said to be included within the scope of the invention of the present application.

In addition, components having the same function within the scope of the same idea shown in the drawings of each embodiment will be described using the same reference numerals.

Figure 1 is a block diagram showing the overall configuration of an agricultural product demand prediction system according to an embodiment of the present invention.

Referring to FIG. 1, a data collection unit that collects relevant data necessary for predicting demand for agricultural products distributors from agricultural product distributors through a predetermined management server, and among the items, quantities and margins of the data collected by the data collection unit. A database unit that stores data containing at least one information by item, an analysis unit that analyzes the item, the quantity, and the margin of agricultural products based on the data by item stored in the database unit, and the analysis unit analyzes the product, An item forecasting unit that predicts high-order items above a preset value according to a preset calculation method for expected demand items, and predicts the quantity of the expected demand quantity of the item predicted by the item forecasting unit within a preset range according to a preset calculation method. A quantity prediction unit that assumes sales of the quantity of the item predicted in the quantity prediction unit, and a margin prediction unit that predicts the seller's sales margin within a preset range according to a predetermined calculation method, and the item prediction unit predict It may include a proposal unit that finally calculates the margin of the margin prediction unit for the item and the quantity predicted in the quantity prediction unit and proposes a sales quantity to the seller.

Figure 2 is a configuration diagram showing an agricultural product demand prediction system according to an embodiment of the present invention.

Referring to Figure 2, it is possible to predict demand for products/services during a specific period through machine learning. A demand forecasting system learned based on past data increases prediction and analysis accuracy. Unlike existing demand forecasting methods, machine learning technology (accelerates data processing, improves forecast accuracy, automatically updates demand forecast based on recent data, allows analysis based on more data, makes it easier to identify patterns in data, builds a solid system, makes unpredictable It has the advantage of improved adaptability to change. This technical method allows for more accurate and reliable predictions by classifying data in advance. Demand forecasting results vary depending on product type and product category. Short-term forecasting is generally done for less than 12 months (1 week/1 month/6 months), but long-term forecasting takes a period of more than a year.

When building a demand forecasting model, data based on (Consistency, Accuracy, Validity, Relevance, Accessibility, Completeness, Detalization) must be used. Companies generally need to organize the collected data, analyze irregular data, and identify relationships to classify the data. Using valid data obtained by classifying data, relationships are identified and structured in a comprehensive form. There is no single forecasting algorithm that fits all situations, but a demand forecasting model tailored to each situation can be built using various machine learning methods determined by factors such as business goals, data type, data quantity and quality, and forecast period. there is. The demand forecasting program is conducted using the following methods (ARIMA/SARIMA, Linear Regression, XGBoost, Random Forest, Long Short Term Memory LSTM, Moving Average).

After a demand prediction model is built, a process of learning data is necessary. Primarily processing historical data, the algorithm provides usable predictions.

For a high-accuracy forecasting algorithm, the parameters of the demand forecasting model must be optimized. Divide the learning data into 10 equal parts and cross-verify each demand forecasting program. Parameters are tested precisely and prediction algorithms are trained.

Machine learning makes it possible to predict demand for products/services over a specific period of time. A demand forecasting system learned based on past data increases prediction and analysis accuracy. Unlike existing demand forecasting methods, machine learning technology (accelerates data processing, improves forecast accuracy, automatically updates demand forecast based on recent data, allows analysis based on more data, makes it easier to identify patterns in data, builds a solid system, makes unpredictable It has the advantage of improved adaptability to change.

This technical method allows for more accurate and reliable predictions by classifying data in advance. Before going into technical details, it is a good idea to establish the purpose of the output data our company wants to obtain. Demand forecasting results vary depending on product type and product category. Short-term forecasting is generally done for less than 12 months (1 week/1 month/6 months), but long-term forecasting takes a period of more than a year.

When building a demand forecasting model, data based on consistency, accuracy, validity, relevance, accessibility, completeness, and detalization (Consistency, Accuracy, Validity, Relevance, Accessibility, Completeness, Detalization) must be used. Companies generally need to organize the collected data, analyze irregular data, and identify relationships to classify the data. When developing POS programs for retail customers, utilize technology to ensure high quality and data stability. Using valid data obtained by classifying data, relationships are identified and structured in a comprehensive form. There is no single forecasting algorithm that fits all situations, but a demand forecasting model tailored to each situation can be built using various machine learning methods determined by factors such as business goals, data type, data quantity and quality, and forecast period. there is. The demand forecasting program is conducted using the following methods (ARIMA/SARIMA, Linear Regression, XGBoost, Random Forest, Long Short Term Memory LSTM, Moving Average).

After a demand prediction model is built, a process of learning data is necessary. Primarily processing historical data, the algorithm provides usable predictions. Provides a trained model.

For a high-accuracy forecasting algorithm, the parameters of the demand forecasting model must be optimized. Divide the learning data into 10 equal parts and cross-verify each demand forecasting program. Parameters are tested precisely and prediction algorithms are trained.

Meanwhile, according to one embodiment of the present invention, the entire ordering process is as follows.

(a) Orders are accepted at Mr. Papa’s store or online store.

(b) If the ordered product is available in the warehouse, it will be delivered to the partner carrier.

(c) Check the product inventory and according to the quantity of the product, the company purchases the product from the farmer.

(d) There are two methods: securing it directly from the farmer and winning it through an auction.

To implement this, the currently available applications are as follows.

(a) Optimal route finding program to reduce shipping costs

(b) Sales analysis program (AI) to analyze the same sales in real time

(c) a warehouse management system that includes algorithms that use barcode or product code scanning to locate products, assign new locations when space for the same product becomes full, and store optimal product locations;

(d) an order management system (ERP) that manages online or offline orders, downloads data for order managers, sorts orders, and manages inventory regardless of whether the product is in the warehouse or not;

The above agricultural product demand forecasting system is E-commerce forecasting is the process of estimating future demand for products. And the final forecast is typically based on historical metrics, such as historical sales data, and real-time inventory trends, such as current inventory levels.

And the hardest part is that you don't know how much product you will need in the next few days, so you have to build AI to predict sales for the next day.

According to AI's predictions, we can buy products from farmers, in which case we will reduce expired products because AI will predict the future of sales through order history.

The customers of the agricultural product demand forecasting system are as follows.

Access key customer data and understand the value and frequency of shopping behavior.

In many ways, inventory forecasting is key to customer satisfaction, so forecasting allows you to create replenishment forecasts that accurately reflect product demand and customer purchasing patterns. This way, you can ensure that you always have products in stock and ready to ship when your customers want them.

The inventory forecast of the agricultural product demand forecasting system is as follows.

It is possible to regularly compare seasonal trends and visualize performance.

Orders and inventory levels are processed based on ERP, but in many cases these systems used incomplete historical data to predict quantities needed to replenish inventory. This resulted in high wasted costs and low service levels for many products. We want to predict the optimal order quantity using a data-based model. The goal of this project is to support ordering decisions by creating a user-friendly tool to determine optimal sizes and reorder levels.

The cost savings of the agricultural product demand forecasting system are as follows.

Another important benefit of demand forecasting is that it can reduce overhead costs.

Forecasting allows companies to order exactly what they need to meet demand, ensuring they don't overspend on products that may end up in excess inventory.

And avoiding excess inventory on low-selling products actually prevents brands from paying additional shipping costs.

Saving money through demand forecasting means you have more working capital available to invest elsewhere in your business.

Meanwhile, the real-time auction tracking system determines the price through radio in real time during an auction in a fresh product auction market across the country, enabling tracking and purchasing of the lowest price product, and being able to respond differently from other companies in terms of price competitiveness.

The optimal route finding program of the agricultural product demand prediction system is an application that provides a solution for finding the optimal route for drivers.

Services we provide to get a better route to the customer location and services that help us manage driver location services while providing classification of each distributed order according to the customer's zone and help avoid traffic problems on the road This can include authentication to help ensure safe and secure data for services and applications, such as mapping services that help you find the optimal and secure location.

Additionally, data collection programs include services that store large amounts of data on products, services that manage and change every detail of product data, and services that sort and filter large amounts of data at the same time, as well as services that automatically classify products inside the application. Product recognition, code entry, and big-size data can be downloaded to the database in real time and inserted for data filtering.

The Analytical Real-Time Application of the Agricultural Demand Forecasting System is an application with PapaMarket's branches and services that provide better visualization and explanation of data.

The functional service shows the location of each branch and warehouse of Papamarket on a map.

It may include services that help you present real-time budget objectives for an accurate period of time, services that help you track real-time sales data for each branch and warehouse in your market, and services that help you keep track of your branch listings. .

Meanwhile, supply chain innovation and automation for drivers of the agricultural demand forecasting system is key to knowing what we own in real time.

Integration gives us near real-time visibility into where our assets are within the fleet to ensure cargo arrives at the right location on time.

Stores can predict cargo arrival times and effectively plan work and labor for truck deliveries. Customers can also benefit from this increased visibility and efficiency.

They will be able to see our shelves stocked in stores and have access to a more accurate inventory of products available for purchase online.

It allows drivers to communicate more closely with stores by integrating with store applications to provide a smoother delivery process. As a driver approaches the store, location tracking technology detects the driver's location and sends a push notification to store associates' POS devices so they can quickly plan unloading and turnaround. You can also receive important information from the scene via walkie-talkie with the driver.

One of the goals is to create a frictionless workflow so drivers can spend more time driving trucks and less time waiting in warehouses or stores.

All systems are fully integrated to coordinate schedules and navigation, allowing drivers to conveniently navigate to their next destination. As with any job, unplanned activities sometimes require drivers to do more than steer a truck. Our new system creates accountability by allowing drivers to communicate their achievements on the road.

The warehouse operation system of the above agricultural product demand forecasting system is such that every warehouse worker has his or her own account, and the manager can view the work performed to check the worker's performance and mistakes.

Figure 3 is a configuration diagram showing the sales structure of an agricultural product demand prediction system according to an embodiment of the present invention.

Referring to Figure 3, the agricultural product demand forecasting system collects demand data for Mr. Dad products, learns demand data, classifies past data of Mr. Dad products and stores, adds specific dates to high-demand products, and builds a demand prediction model. By analyzing, you can predict the demand for Mr. Dad products based on a specific date and visualize the forecast graph.

More specifically, in the data collection stage, data on product name, store name, sales date, and sales volume can be collected.

Afterwards, in the analysis stage, after tracking accurate store information and demand for each product, learning market demand history, organizing Mr. Dad Market's past data, selecting the optimal date to sell products within an accurate period, adding specific dates to high-demand products, and This can include adding the most sold products on a date, building demand forecasting models, and visualizing graphs.

Finally, in the demand forecasting stage, it is possible to forecast demand for each store, forecast demand for an accurate period, forecast accurate product demand, and compare historical and forecast data for product demand.

Meanwhile, the subdivision system that constitutes the agricultural product demand prediction system may include the following steps.

(a) On/offline real-time demand status counting

(b) Advance demand forecasting system demand status & past sales data analysis

(c) Real-time confirmation available at the production/subdivision center

(d) Arrival of goods from the production site to the raw material distribution center

(e) Delivery to distribution center after small distribution packaging

4 to 17 are diagrams illustrating a demand prediction model of the agricultural product demand prediction system according to an embodiment of the present invention.

With reference to FIGS. 4 to 17, an example of the agricultural product demand prediction system is as follows.

[Example 1]

This consists of comprehensive Exploratory Data Analysis with Prophet and MLP Neural Network Forecast Modeling for Store Item Demand Forecasting. The objective of the problem is to forecast 3 months of sales for 50 different items at 10 different stores using the 5 years history of sales.

The data contains the following files:

train.csv: the training data with the history of sales. 913,000 data from 2013 to 2017.

test.csv: the test data to predict the sales. 44,999 data in 2018.

prediction.csv: the output data which is predicted by the AI

Data fields:

date - Date of the sale data. There are no holiday effects or store closures.

store - Store ID

item - Item ID

sales - Number of items sold at a particular store on a particular date.

Time series forecasting is a skill that few people claim to know. Machine learning is cool. And there are a lot of people interested in becoming a machine learning expert. But forecasting is something that is a little domain specific.

Retailers like Walmart, Target use forecasting systems and tools to replenish their products in the stores. An excellent forecast system helps in winning the other pipelines of the supply chain. If you are good at predicting the sale of items in the store, you can plan your inventory count well. You can plan your arrangement well.

A good forecast leads to a series of wins in the other pipelines in the supply chain.

A time series is a sequence of observations collected at some time intervals. Time plays an important role here. The observations collected are dependent on the time at which it is collected.

The sale of an item say bread in a retail store like Walmart will be a time series. The sale could be at daily level or hourly level. Time is important here. During Chusok holidays, this number would be humongous compared to the other days. This is known as seasonality.

Frequency

When setting the frequency, many people are confused what should be the correct value. This is the simple definition of frequency. Frequency is the number of observations per cycle. Now, how you define what a cycle is for a time series?

Say, you have electricity consumption in Seoul at hourly level. The cycle could be a day, a week or even annual. I will cover what frequency would be for all different types of time series.

Before we proceed I will reiterate this.

Frequency is the number of observations per cycle. As we want to analyze the daily demand, there could be weekly or annual cycle, 7 or 365(366) respectively.

Some of the years have 366 days (leap years). So if your time series data has longer periods, it is better to use frequency = 365.25. This takes care of the leap year as well which may come in your data.

Frequency Types

Weekly data) There could be an annual cycle. frequency = 52 and if you want to take care of leap years then use frequency = 365.25/7

Monthly data) Cycle is of one year. So frequency = 12

Quarterly data) Again cycle is of one year. So frequency = 4

(Yearly data) Frequency = 1

There could be smaller interval frequency as well like*?*Hourly, Half-Hourly, Minutes, Seconds time series data as well

Used Frameworks and Libraries

Sklearn

Scipy

Plotly

pandas

lightgbm

xgboost

Seaborn

Fastai

Fbprophet

Sample Input Data Structure

Date => time series

Store => id number of the store

Item => id number of the product

Sales => number of sold products in that specific time series

Moving Average

Let's start with a naive hypothesis: "tomorrow will be the same as today". However, instead of a model like yt=yt*?*1 (which is actually a great baseline for any time series prediction problems and sometimes is impossible to beat), we will assume that the future value of our variable depends on the average of its k previous values. Therefore, we will use the moving average.

Exponential smoothing

Now, let's see what happens if, instead of weighting the last k values of the time series, we start weighting all available observations while exponentially decreasing the weights as we move further back in time. There exists a formula for exponential smoothing that will help us with this:

Here the model value is a weighted average between the current true value and the previous model values. The α weight is called a smoothing factor. It defines how quickly we will “forget” the last available true observation. The smaller α is, the more influence the previous observations have and the smoother the series is.

Exponentially is hidden in the recursiveness of the function -- we multiply by (1*?*α) each time, which already contains a multiplication by (1*?*α) of previous model values.

Double exponential smoothing

Up to now, the methods that we've seen are for a single future point prediction (with some nice smoothing). That is cool, but it is also not enough. Let's extend exponential smoothing so that we can predict two future points (of course, we will also include more smoothing).

Series decomposition will help us -- we obtain two components: intercept (i.e. level) ℓ and slope (i.e. trend) b. We have learned to predict intercept (or expected series value) with our previous methods; now, we will apply the same exponential smoothing to the trend by assuming that the future direction of the time series changes depends on the previous weighted changes. As a result, we get the following set of functions:

The first one describes the intercept, which, as before, depends on the current value of the series.

The second term is now split into previous values of the level and of the trend.

The second function describes the trend, which depends on the level changes at the current step and on the previous value of the trend. In this case, the β coefficient is a weight for exponential smoothing.

The final prediction is the sum of the model values of the intercept and trend.

Modeling Holidays and Special Events

We have to consider the Holidays, because stores will have good sales in special Holidays.

If there are holidays or other recurring events that we'd like to model, we must create a Data Frame for them. It has two columns (holiday and ds) and a row for each occurrence of the holiday. It must include all occurrences of the holiday, both in the past (back as far as the historical data go) and in the future (out as far as the forecast is being made).

If they won't repeat in the future, the Prophet will model them and then not include them in the forecast.

As a result, the relative error of our forecast (MAPE) is about 27.5%, and on average our model is wrong by 3.54 predicts (MAE).

In addition, although the configuration and features of the present invention have been described above based on the embodiments according to the present invention, the present invention is not limited thereto, and various changes or modifications may be made within the spirit and scope of the present invention. It is obvious to those skilled in the art, and therefore, it is stated that such changes or modifications fall within the scope of the appended patent claims.

110: Data collection department
120: database unit
130: analysis department
140: Item prediction unit
150: Quantity forecasting department
160: Margin prediction unit
170: Proposal section

Claims (1)

A data collection unit that collects relevant data necessary for predicting demand for agricultural products distribution companies from agricultural products distribution companies through a predetermined management server;
a database unit that stores data containing at least one of the item, quantity, and margin of the data collected by the data collection unit by item;
an analysis unit that analyzes the item, the quantity, and the margin of agricultural products based on item-specific data stored in the database unit;
an item prediction unit that predicts higher items than a preset value based on a predetermined calculation method for the expected demand items analyzed in the analysis unit;
a quantity prediction unit that predicts the expected demand quantity of the item predicted by the item prediction unit within a preset range according to a predetermined calculation method;
a margin prediction unit that assumes sales of the quantity of the item predicted by the quantity prediction unit and predicts the seller's sales margin within a preset range according to a predetermined calculation method; and
a proposal unit that finally calculates the margin of the margin prediction unit for the item predicted in the item prediction unit and the quantity predicted in the quantity prediction unit and proposes a sales quantity to the seller;
An agricultural product demand forecasting system comprising: