Part demand prediction method and device, computer equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a part demand prediction method and device, computer equipment and a storage medium, and the method comprises the steps: classifying parts according to the life cycle and order quantity of the parts, and obtaining the types of the parts; constructing a part label system according to the types of the parts; preprocessing the part label data according to the part label system; according to the preprocessed part label data, constructing part demand prediction features; constructing a prediction model according to the part demand prediction features; and according to the prediction model, predicting the part demand. The reliability is high, multiple types of models are fused and complement each other, and the reliability of a prediction result is improved; attribute classification is carried out on the parts, different prediction methods are provided, the influence of abnormal orders on model prediction is effectively solved, and the overall prediction effect is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a parts demand forecasting method, device, computer equipment and storage medium. Background Art

[0002] With the intensified competition in the automotive industry and the rapid advancement of technology, how to enhance market competitiveness, meet customer delivery, reasonably control inventory, increase cash flow, and achieve a balance between demand and inventory has become a challenge faced by every automotive parts department. Accurate forecasting of parts demand is the basis of production planning and inventory management, and is crucial to the operation of the entire supply chain. Accurate demand forecasting can help companies avoid the risk of excess or shortage of parts inventory, thereby optimizing resource allocation, reducing costs, and enhancing the company's market competitiveness and response speed.

[0003] As a key link in the supply chain management of enterprises, the accuracy of parts demand forecasting directly affects inventory control, production planning and cost-effectiveness. For example, the demand forecast of automotive parts plays an extremely important role in supply chain management, which directly affects the operational efficiency, cost control and customer satisfaction of enterprises. The demand forecast of parts is mainly based on the demand judgment model for the future combined with the historical demand of parts. In the process of ordering parts, there are certain differences in the corresponding demand under different life cycles and different order quantities. Therefore, the forecast of parts must be based on multiple conditions for analysis to be more reliable. The existing technology generally focuses on demand forecasting for a single attribute, which is of low practicality. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for forecasting demand for parts, which adopts the following technical solution, including:

[0005] Classify the parts according to their life cycle and order quantity to obtain the categories of the parts;

[0006] Build a parts labeling system based on parts categories;

[0007] Pre-process the parts label data according to the parts label system;

[0008] Construct component demand prediction features based on preprocessed component label data;

[0009] Constructing a prediction model according to the component demand prediction characteristics;

[0010] The demand for the components is predicted based on the prediction model.

[0011] Preferably, the step of classifying the parts according to the life cycle and order quantity of the parts to obtain the categories of the parts specifically includes:

[0012] Collect historical spare parts demand data for components;

[0013] Based on historical spare parts demand data, formulate classification standards for stable spare parts and unstable spare parts;

[0014] According to the classification standards, spare parts are divided into two categories: stable spare parts and unstable spare parts.

[0015] Preferably, the step of constructing a component labeling system according to the categories of components specifically includes:

[0016] Set label classification criteria;

[0017] According to the label classification standards, set the correspondence between label categories and label attributes to build a component label system.

[0018] Preferably, the step of preprocessing the component label data according to the component label system specifically includes:

[0019] According to the parts labeling system, clean the parts labeling data;

[0020] Merge and integrate cleaned component label data from multiple data sources to ensure data consistency and coherence;

[0021] Data reduction is performed on the merged and integrated component label data.

[0022] Preferably, the step of constructing component demand prediction features based on the preprocessed component label data specifically includes:

[0023] Establish an initial feature set and filter features;

[0024] Encode categorical features and convert them into numerical features;

[0025] Scale numerical features to make different features comparable in value;

[0026] According to business needs, multiple features are combined to form new features.

[0027] Preferably, the step of constructing a prediction model according to the component demand prediction characteristics specifically includes:

[0028] Select a machine learning model based on the characteristics of parts demand data and business needs;

[0029] Train machine learning models;

[0030] Validate machine learning models.

[0031] Preferably, the step of predicting the demand for parts according to the prediction model specifically includes:

[0032] Based on the forecasting model, identifying internal and external factors that affect parts demand;

[0033] Assess the extent to which internal and external factors affect the demand for parts;

[0034] Based on the degree of impact, the demand for the components is predicted.

[0035] In order to solve the above technical problems, the present invention also provides a component demand forecasting device, which adopts the following technical solution, including:

[0036] A classification module is used to classify parts according to their life cycle and order quantity to obtain the categories of parts;

[0037] The first building module is used to build a parts labeling system according to the categories of parts;

[0038] A preprocessing module is used to preprocess the component label data according to the component label system;

[0039] The second construction module is used to construct component demand prediction features based on the preprocessed component label data;

[0040] A third building module is used to build a prediction model according to the component demand prediction characteristics;

[0041] The prediction module is used to predict the demand for the parts according to the prediction model.

[0042] In order to solve the above technical problems, the present invention also provides a computer device, which adopts the technical solution described below, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the above-mentioned component demand forecasting method when executing the computer-readable instructions.

[0043] In order to solve the above technical problems, the present invention also provides a computer-readable storage medium, which adopts the technical solution described below, and stores computer-readable instructions on the computer-readable storage medium. When the computer-readable instructions are executed by the processor, the steps of the above-mentioned component demand forecasting method are implemented.

[0044] Compared with the prior art, the present invention has the following beneficial effects: first, by classifying parts according to their life cycle and order quantity, enterprises can clearly distinguish different categories of parts such as high demand, medium demand and low demand. This classification helps enterprises identify key parts and thus prioritize resource allocation; second, a parts labeling system is constructed to systematically organize parts information, including model, purpose, supplier, etc., to provide a basis for subsequent data processing and help achieve standardization and unified management of parts information; then, the parts label data is preprocessed, such as deduplication, cleaning and formatting, to ensure data quality and reduce prediction deviations caused by data errors; then, based on the preprocessed data, parts demand prediction features are constructed, such as historical demand, seasonal fluctuations, market trends, etc., to more comprehensively reflect the inherent laws of parts demand; these features are then used to construct a prediction model, For example, time series analysis and machine learning algorithms can achieve accurate prediction of parts demand. These models can capture changes in demand and improve the timeliness and accuracy of predictions. Finally, by predicting parts demand based on the prediction model, companies can adjust production plans and optimize inventory levels in advance to cope with changes in market demand. This helps reduce inventory costs, improve production efficiency, and enhance the company's market competitiveness. It has strong reliability and integrates multiple models to complement each other, increasing the reliability of prediction results. It classifies parts by attributes and proposes different prediction methods, which effectively solves the impact of abnormal orders on model predictions and improves the overall prediction effect. Taking into account the actual application scenarios, the framework of some prediction models supports the identification and interpretation of trends, seasonality and other factors. For machine learning models, model interpretation algorithms and tools can be used to identify feature importance and important influencing factors of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the scheme in the present invention, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 is a flow chart of an embodiment of the parts demand forecasting method of the present invention;

[0047] Figure 2 is an architectural diagram of an embodiment of a method for forecasting component demand of the present invention;

[0048] Figure 3 It is a structural schematic diagram of an embodiment of the component demand forecasting device of the present invention;

[0049] Figure 4It is a schematic structural diagram of an embodiment of a computer device of the present invention. DETAILED DESCRIPTION

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present invention; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present invention or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0051] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0052] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0053] It should be noted that the parts demand forecasting method provided in the embodiment of the present invention is generally executed by a server / terminal device, and accordingly, the parts demand forecasting device is generally arranged in the server / terminal device.

[0054] It should be understood that the number of terminal devices, networks and servers is only illustrative. Depending on the implementation requirements, there can be any number of terminal devices, networks and servers.

[0055] Embodiment 1

[0056] Please refer to Figure 1 and Figure 2 , shows a flow chart of an embodiment of the component demand forecasting method of the present invention and an architecture diagram of an embodiment of the component demand forecasting method of the present invention. The component demand forecasting method comprises the following steps:

[0057] Step S1, classifying the parts according to their life cycle and order quantity to obtain the categories of the parts.

[0058] In this embodiment, the electronic device (e.g., server / terminal device) on which the component demand forecasting method runs can receive the component demand forecasting request via a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAXX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0059] In this embodiment, step S1, classifying the parts according to the life cycle and order quantity of the parts, and obtaining the categories of the parts may further include the following steps:

[0060] S11, collect historical spare parts demand data of components.

[0061] The Enterprise Resource Planning (ERP) system contains information on the purchase, inventory, use and scrapping of spare parts. Through the ERP system, you can query the spare parts purchase records in the past period of time, and thus obtain historical spare parts demand data.

[0062] S12. Based on the historical spare parts demand data, formulate classification standards for stable spare parts and unstable spare parts.

[0063] Stable spare parts refer to spare parts with relatively stable demand and small fluctuations; while unstable spare parts refer to spare parts with large fluctuations in demand and difficult to predict.

[0064] Based on the historical spare parts demand data, a basic classification method standard can be developed to classify parts.

[0065] The basic classification method standard refers to the classification of parts based on the median monthly order sales volume as the classification indicator.

[0066] S13, according to the classification standards, the spare parts are divided into two categories: stable spare parts and unstable spare parts.

[0067] According to the basic classification method standards, based on the historical demand for each spare parts, for example, the median monthly sales volume of orders in the past 36 months is used as the classification indicator, the spare parts will be divided into five categories: A, B, C, D and E.

[0068] Unstable spare parts (Class E): Considering that the life cycle of spare parts changes with the status of the vehicle after it is launched, it is divided into three stages: the new part stage, the production stage and the discontinued stage. The sales of spare parts in the new part and discontinued stages are unstable. Such parts usually show that there are too many months with zero sales within (36 months), and the sales trend is not enough for the title model to predict future sales. Generally, the time series method is used for prediction. The division rule here takes 36 months of data as an example. If the number of months with zero sales data is greater than or equal to 24, that is, the number of months with sales in the training data does not meet 12 months, the parts are unstable.

[0069] Stable spare parts (A, B, C and D): Stable parts are spare parts with non-zero sales data for more than 12 months. Stable parts are divided into 90%, 5%, 3% and 2% based on historical order demand.

[0070] It should be noted that in addition to using the basic classification method standard, the cluster classification method can also be used to divide the parts through the coefficient of variation (CV) and the average demand interval (ADI).

[0071] The clustering classification method uses the K-shape algorithm to calculate the similarity between different sequences. It uses a new distance definition method, which enables K-Shape to do both subsequence clustering (extracting subsequences through sliding windows, that is, clustering fragments in a single long time series, which can effectively discover frequent patterns and abnormal fragments in longer time series) and full sequence clustering.

[0072] Spare parts are divided into four categories by coefficient of variation (CV) and average demand interval (ADI), for example by percentile:

[0073] Greater than CV75 quantile Less than or equal to CV75 quantile Greater than ADI 75th percentile A C Less than or equal to ADI 75th percentile B D

[0074] Here, the coefficient of variation (CV) is used to measure the change in demand over time. The larger the CV value, the more drastic the fluctuation in demand. CV = standard deviation of demand / mean demand. The average demand interval (ADI) is used to measure the average level of the interval between time windows with non-zero demand. ADI = total number of periods / total number of periods with demand.

[0075] Step S2, constructing a component labeling system according to the categories of components.

[0076] In this embodiment, step S2, building a component label system according to the category of the component may further include the following steps:

[0077] S21, setting label classification standards.

[0078] Set label classification standards, including basic information, policy impulse, replacement parts, specialty store orders, BOM, spare parts sales, holidays, vehicle sales, vehicle maintenance and other classification standards.

[0079] S22, according to the label classification standard, set the corresponding relationship between the label category and the label attribute, and build a component label system.

[0080] Under different label categories, you need to set label attributes corresponding to these label categories. The label categories and their label attributes constitute the component label system. For example, the component label system can be shown in the following table:

[0081]

[0082]

[0083] Step S3, pre-processing the component label data according to the component label system.

[0084] In the forecast of parts demand, parts label data is an important source of information. Parts label data includes the part code, name, specification, supplier information, production date, etc. In order to ensure the accuracy of the forecast, these label data need to be preprocessed.

[0085] In this embodiment, step S3, preprocessing the component label data according to the component label system may further include the following steps:

[0086] S31, performing data cleaning on the component label data according to the component label system.

[0087] For component label data, outliers may be affected by policies, environment and other factors, do not contain meaningful information, and will affect model learning and prediction of future trends.

[0088] Abnormal data can be divided into: missing data, inf or none values; invalid data, which is affected by policies or environment and shows a sudden surge or drop compared with history. Invalid values ​​are filtered out by combining isolation forests, and the average of the sales values ​​of the adjacent months is used for missing data and invalid values.

[0089] Use a function or algorithm to identify missing values ​​in the data. For example, in R, you can use the `is.na` function to identify missing values. Missing values ​​can be handled by deletion, replacement, or interpolation, depending on the case.

[0090] When missing values ​​have little impact on the research objectives, we can directly delete the observations or variables with missing values. In R language, we can use the `na.omit` function to delete rows with missing values.

[0091] For numeric variables, missing values ​​can be replaced by the mean of other numbers under the variable; for non-numeric variables, missing values ​​can be replaced by the median or mode.

[0092] Missing values ​​are interpolated through regression models or multiple imputation algorithms. Regression imputation uses the interpolated variable as the dependent variable and other variables as independent variables, and uses the regression model to fit. Multiple imputation generates a complete set of data from a data set containing missing values, interpolates multiple times, and generates a random sample of missing values.

[0093] Identify outliers through univariate scatter plots or box plots. In R, you can use the `dotchart` function to draw univariate scatter plots and the `boxplot` function to draw box plots.

[0094] For the identified outliers, you need to first analyze the possible reasons for their occurrence, and then determine whether they need to be discarded or replaced. If the outliers are caused by data entry errors or measurement errors, they can be deleted or replaced with normal values; if the outliers reflect the actual business situation, they should be retained.

[0095] S32, merging and integrating the cleaned component label data from multiple data sources to ensure data consistency and coherence.

[0096] Merge data from different data sources based on keywords. For example, in R language, you can use the `merge` function to merge two data frames based on keywords.

[0097] Dealing with homonyms and synonyms: During the data integration process, homonyms and synonyms may occur. Homonyms refer to the same attribute names in different data sources, but the entities they represent are different, and they cannot be used as keywords; synonyms refer to different attribute names in different data sources, but the entities they represent are the same, and they can be used as keywords.

[0098] Data integration often results in data redundancy, which may be due to multiple occurrences of the same attribute or duplication caused by inconsistent attribute names. For duplicate attributes, if any, they will be deleted.

[0099] S32, performing data reduction on the merged and integrated component label data.

[0100] Data reduction is to minimize the amount of data while keeping the original appearance of the data as much as possible, so as to reduce the impact of invalid and erroneous data on modeling and save time and space for storing data.

[0101] Attribute reduction, find the smallest attribute subset, and make sure the probability distribution of the subset is close to the probability distribution of the original data. Attribute reduction methods include stepwise selection (starting from an empty attribute set, each time selecting the current worst attribute in the original attribute set and removing it until the worst attribute cannot be selected or a constraint value is satisfied), decision tree induction (removing attributes that do not appear in the decision tree from the initial set to obtain a better attribute subset) and principal component analysis (using fewer variables to explain most of the variables in the original data, and using highly correlated variables to transform them into independent or unrelated variables).

[0102] Numerical reduction reduces storage and computational costs by reducing the amount of data. Numerical reduction methods include parametric methods (such as linear regression and multivariate regression) and non-parametric methods (such as histograms and sampling).

[0103] Step S4, constructing component demand prediction features based on the preprocessed component label data.

[0104] In this embodiment, step S4, constructing the component demand prediction feature according to the preprocessed component label data, may further include the following steps:

[0105] S41, establish an initial feature set and filter features.

[0106] The potential factors that affect the demand for parts are cleaned and preprocessed to form an initial feature set. These potential factors include the historical sales volume, type (new products, seasonal products, promotional products, regular products), region, order quantity, etc. of the parts.

[0107] Recursive feature elimination, cross-validation, principal component analysis, random forest and other methods are used to screen the initial feature set and retain the features that have a significant impact on the prediction results.

[0108] The initial feature set includes basic spare parts features, sales volume features, holiday features, maintenance order data features, new and old parts replacement features, and policy data.

[0109] In practice, 5676 individual spare parts in eight categories of five flow rates with three frequencies, high, medium and low were analyzed:

[0110] The basic features of spare parts can be processed as follows:

[0111] One-hot encoding is used for categorical features, including flow rate, category, supplier, etc. For example, flow rate data is represented by high, medium and low. After One-hot encoding, 100 represents high, 010 represents medium and 001 represents low.

[0112] Numeric variables are used directly, such as order quantity, order frequency, delivery frequency, etc.

[0113] The sales volume characteristics can be processed as follows:

[0114] For aggregate sales, group and aggregate sales by spare part number and month;

[0115] For sales growth rate, sort by part number and month and calculate the sales growth rate for 24 months;

[0116] For periodic sales, the average, median, maximum and minimum values ​​of window sales are calculated with rolling orders of 3, 6, 9, 12, 15 and 18 respectively;

[0117] Month-on-month characteristics include year-on-year characteristics and historical month-on-month characteristics;

[0118] The fluctuation characteristics include 1, 2, 3, 6, 12, and 24 lagged sales, and the growth rate is calculated;

[0119] Seasonal characteristics include end-of-quarter impulse, end-of-quarter sales trend, and quarterly sales average.

[0120] The following processing is done for holiday features:

[0121] One-Hot coding is used for statutory holidays; data statistics are used as features for working days and rest days.

[0122] Maintenance order data may be ahead of or behind sales data, and the overall cyclical trend is close to sales data. Therefore, feature construction includes: a. Cyclic features: the average, median, maximum and minimum values ​​of window sales are calculated with rolling orders 3, 6, 9, 12, 15 and 18 respectively; b. Seasonal features, including end-of-quarter impulse, end-of-quarter sales trend and quarterly sales average; c. Month-on-month features, including year-on-year features and historical month-on-month features.

[0123] The replacement of old parts with new ones is manifested in a decrease in the demand for old parts and an increase in the demand for replacement parts. Here, the month of the intersection of the sales volume of new and old parts (new parts demand >= old parts demand) is taken, and the historical demand data of old parts before the intersection is added to the historical demand data of new parts as a supplement.

[0124] Policy data generally include recalls or promotions, etc. Based on the given activity start time and end time, the spare parts activity characteristics that fall within the activity cycle are set to 1, otherwise they are set to 0.

[0125] S42, encoding the categorical features and converting the categorical features into numerical features.

[0126] Encode categorical features, such as using One-Hot Encoding or Label Encoding, to convert categorical features into numerical features.

[0127] S43, scales the numerical features to make different features comparable in value.

[0128] Scaling numerical features, such as using standardization or normalization methods, makes different features comparable in value.

[0129] S44, combining multiple features to form new features according to business needs.

[0130] By adopting step S44, more complex patterns can be captured.

[0131] Step S5, constructing a prediction model based on the component demand prediction characteristics.

[0132] The prediction model can be a statistical model, a machine learning model, a deep learning model, etc. When selecting a model, factors such as the characteristics of the component demand data, the accuracy of the prediction, and the complexity of the model need to be considered.

[0133] In this embodiment, step S5, building a prediction model according to the component demand prediction characteristics may further include the following steps:

[0134] S51, select a machine learning model based on the characteristics of the parts demand data and business needs.

[0135] For high-end and medium-end parts, you can select high-end and medium-end prediction models. High-end and medium-end flow prediction models can include time series models, machine learning models, and deep learning models.

[0136] The time series model mainly consists of three parts: growth trend, seasonal trend and holiday impact:

[0137] y(t)=g(t)+s(t)+h(t)+e(t).

[0138] Among them, g(t) represents the trend term, which indicates the non-periodic trend of the time series. s(t) represents the periodic term, or seasonal term, which is generally expressed in weeks or years. h(t) represents the holiday term, which indicates the impact of potential holidays with non-fixed periods in the time series on the predicted value. e(t) is the error term or residual term, which indicates the fluctuations not predicted by the model and obeys the Gaussian distribution.

[0139] Machine learning models, including the latest versions of XGBoost, LightGBM, and CatBoost, are suitable for different scenarios. And due to its excellent code structure and API design, it is convenient to call, train, and test models in a consistent and efficient way in applications. In addition, for specific spare parts types, you can choose Random Forest and its derivative algorithm Extreme Random Forest (Extreme RF) to reduce the impact of bias and overfitting.

[0140] Deep learning algorithms: In terms of deep learning methods, combined with the automotive spare parts scenario, we can use the time series prediction deep learning algorithm LSTM and deep learning algorithms based on Transformer and other frameworks. Here, the Informer and TFT (Temporal Fusion Transformer) models based on the Transformer framework are used. In addition to having good prediction accuracy, they can also provide a certain degree of explanation and probability estimation for model predictions based on the Attention mechanism.

[0141] For low-flow parts, you can choose a low-flow forecast model. The sales trend of low-flow parts is intermittent. Here, Croston's method is used to separate the actual intervals of sales demand of 0 and non-0, and the exponential smoothing method is used for forecasting.

[0142] Z(t)=ɑ*X(t)+(1-ɑ)*Z(t-1);

[0143] P(t)=β×Q+(1-β)*P(t-1);

[0144] M(t)=Z(t) / P(t);

[0145] Among them, X(t) represents the actual demand in period t, Z(t) refers to the predicted value of demand in period t, P(t) refers to the estimate of the interval in period t, α and β are the corresponding smoothing exponents, and Q represents the number of periods since the last actual non-zero demand.

[0146] You can also choose between recursive forecasting and direct forecasting models based on the characteristics of parts demand data and business needs.

[0147] The advantage of recursive prediction is that the trend of the time series can be integrated into the input process to continuously correct the algorithm, but this method consumes a lot of resources.

[0148] Direct forecasting is the result of forecasting multiple periods at once, ignoring the trend item. For forecast periods within three months, recursive forecasting is used, and for forecast periods longer than three months, direct forecasting can be used.

[0149] You can also select an integrated learning prediction model based on the characteristics of the parts demand data and business needs.

[0150] Ensemble learning uses a stacking method, which means training a model to combine other models. Through stacking, the different characteristics and advantages of the basic models can be fully utilized, thereby improving the performance of the overall model.

[0151] S52, training the machine learning model.

[0152] The parts demand data set is divided into training set, validation set and test set. The training set is used for the preliminary training of the machine learning model, the validation set is used to evaluate the training effect and adjust the machine learning model parameters, and the test set is used to finally judge the performance of the machine learning model.

[0153] During the model training phase, select an appropriate training method based on the nature of the problem, such as supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, transfer learning, or deep learning.

[0154] S53, verify the machine learning model.

[0155] Cross-validation can be used to validate machine learning models. By dividing the dataset into multiple subsets, using different subsets as training sets and validation sets each time, repeatedly training and evaluating the model. This ensures that each data point has a chance to be used as a validation set, thereby improving the stability and reliability of the model. Among them, K-fold cross-validation is a mainstream method, which divides the dataset into K equal-sized parts, selects one part each time as the validation set, and the rest as the training set.

[0156] You can also use the leave-one-out method to validate the machine learning model. Each time, only one data point is left as the validation set, and the rest of the data points are used as the training set. This method is suitable for situations where the data set is small.

[0157] The Hold-Out method can also be used to validate the machine learning model. The entire data set is randomly split into a training set and a test set, usually 70% to 80% of the data is used for training and the rest is used for testing. This method is simple and fast.

[0158] It should be noted that when performing model validation, it is also necessary to select appropriate evaluation indicators. For classification problems, evaluation indicators include accuracy, confusion matrix, precision, recall, and F1 score. These indicators can help to fully understand the performance of the model in different aspects.

[0159] Step S6: forecasting the demand for parts based on the forecasting model.

[0160] In this embodiment, step S6, predicting the demand for parts according to the prediction model may further include the following steps:

[0161] S61, based on the forecasting model, identify the internal and external factors that affect the demand for parts.

[0162] In the forecasting process, some internal and external factors are unpredictable, such as natural disasters, policy changes, market emergencies, etc. These factors may have a significant impact on the demand for parts. Therefore, these unpredictable factors need to be considered in the forecast results and corresponding response strategies need to be formulated.

[0163] S62, assess the extent to which internal and external factors influence the demand for parts.

[0164] Conduct risk assessments of internal and external factors that impact component demand.

[0165] S63, forecast the demand for parts based on the degree of impact.

[0166] Based on the potential internal and external factors that may affect the demand for parts and components and the results of risk assessment, formulate response strategies, such as increasing inventory, adjusting production plans, etc. to predict the demand for parts and components.

[0167] After considering all predictable and unpredictable factors, it is necessary to make a comprehensive judgment on the forecast results of parts demand and draw a final conclusion. This step requires comprehensive consideration of factors such as the accuracy, reliability, and feasibility of the forecast results to provide strong support for the company's decision-making.

[0168] The application of parts demand forecast results is the ultimate goal of parts demand forecasting. Parts demand forecast results can be applied to production planning, inventory management, supply chain optimization, etc. At the same time, it is necessary to continuously monitor the parts demand forecast results in order to detect and adjust the forecast errors in a timely manner.

[0169] The implementation of this embodiment has the following beneficial effects: first, by classifying parts according to their life cycle and order volume, enterprises can clearly distinguish different categories of parts such as high demand, medium demand and low demand. This classification helps enterprises identify key parts and thus prioritize resource allocation; secondly, a parts labeling system is constructed, which can systematically organize parts information, including model, purpose, supplier, etc., to provide a basis for subsequent data processing and help achieve standardization and unified management of parts information; then, the parts label data is preprocessed, such as deduplication, cleaning and formatting, to ensure data quality and reduce prediction deviations caused by data errors; then, based on the preprocessed data, parts demand prediction features are constructed, such as historical demand, seasonal fluctuations, market trends, etc., which can more comprehensively reflect the inherent laws of parts demand; then, these features are used to construct prediction models, such as time series analysis, machine learning algorithms, etc., to achieve accurate prediction of parts demand, which can capture demand changes and improve the timeliness and accuracy of predictions; finally, parts demand is predicted based on the prediction model, and enterprises can adjust production plans and optimize inventory levels in advance to cope with changes in market demand. This helps reduce inventory costs, improve production efficiency, and enhance the company's market competitiveness; it has strong reliability and integrates multiple models to complement each other, increasing the reliability of prediction results; it classifies parts by attributes and proposes different prediction methods, effectively solving the impact of abnormal orders on model predictions and improving the overall prediction effect; taking into account actual application scenarios, the framework of some prediction models supports the identification and interpretation of trends, seasonality and other factors. For machine learning models, model interpretation algorithms and tools can be used to identify feature importance and important influencing factors of prediction.

[0170] The present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0171] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0172] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0173] Embodiment 2

[0174] Further references Figure 3 , as a response to the above Figure 1 The present invention provides an embodiment of a component demand forecasting device, and the embodiment of the device is Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0175] like Figure 3 As shown, the component demand forecasting device 70 of this embodiment includes: a classification module 71, a first construction module 72, a preprocessing module 73, a second construction module 74, a third construction module 75 and a forecasting module 76. Among them:

[0176] A classification module 71 is used to classify the parts according to the life cycle and order quantity of the parts to obtain the categories of the parts;

[0177] A first construction module 72 is used to construct a component label system according to the category of the component;

[0178] A preprocessing module 73, used to preprocess the component label data according to the component label system;

[0179] The second construction module 74 is used to construct the component demand prediction feature according to the preprocessed component label data;

[0180] The third construction module 75 is used to construct a prediction model according to the component demand prediction characteristics;

[0181] The prediction module 76 is used to predict the demand for parts based on the prediction model.

[0182] The implementation of this embodiment has the following beneficial effects: high reliability, integration of multiple types of models, taking advantage of each other's strengths and making up for each other's weaknesses, and increasing the reliability of prediction results; attribute classification is performed on parts, and different prediction methods are proposed, which effectively solves the impact of abnormal orders on model predictions and improves the overall prediction effect; considering the actual application scenarios, the framework of some prediction models supports the identification and interpretation of trends, seasonality and other factors. For machine learning models, the importance of features and important influencing factors of prediction can be identified through model interpretation algorithms and tools.

[0183] Embodiment 3

[0184] To solve the above technical problems, the embodiment of the present invention also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0185] The above-mentioned computer device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 8 having components of a memory 81, a processor 82, and a network interface 83, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (ASIC), a programmable gate array (FPGA), a digital processor (DSP), an embedded device, etc.

[0186] The computer device may be a desktop computer, a notebook computer, a PDA, a cloud server, etc. The computer device may interact with the user through a keyboard, a mouse, a remote control, a touch pad, or a voice control device.

[0187] The memory 81 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 81 may be an internal storage unit of the computer device 8, such as a hard disk or memory of the computer device 8. In other embodiments, the memory 81 may also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 8. Of course, the memory 81 may also include both the internal storage unit of the computer device 8 and its external storage device. In this embodiment, the memory 81 is generally used to store the operating system and various application software installed on the computer device 8, such as computer-readable instructions of the component demand forecasting method, etc. In addition, the memory 81 can also be used to temporarily store various types of data that have been output or are to be output.

[0188] In some embodiments, the processor 82 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 82 is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to run the computer-readable instructions stored in the memory 81 or process data, such as running the computer-readable instructions of the component demand forecasting method.

[0189] The network interface 83 may include a wireless network interface or a wired network interface. The network interface 83 is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0190] The implementation of this embodiment has the following beneficial effects: high reliability, integration of multiple types of models, taking advantage of each other's strengths and making up for each other's weaknesses, and increasing the reliability of prediction results; attribute classification is performed on parts, and different prediction methods are proposed, which effectively solves the impact of abnormal orders on model predictions and improves the overall prediction effect; considering the actual application scenarios, the framework of some prediction models supports the identification and interpretation of trends, seasonality and other factors. For machine learning models, the importance of features and important influencing factors of prediction can be identified through model interpretation algorithms and tools.

[0191] Embodiment 4

[0192] The present invention also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the component demand forecasting method as described above.

[0193] The implementation of this embodiment has the following beneficial effects: high reliability, integration of multiple types of models, taking advantage of each other's strengths and making up for each other's weaknesses, and increasing the reliability of prediction results; attribute classification is performed on parts, and different prediction methods are proposed, which effectively solves the impact of abnormal orders on model predictions and improves the overall prediction effect; considering the actual application scenarios, the framework of some prediction models supports the identification and interpretation of trends, seasonality and other factors. For machine learning models, the importance of features and important influencing factors of prediction can be identified through model interpretation algorithms and tools.

[0194] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the various embodiment methods of the present invention.

[0195] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are given in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. A method for forecasting component demand, characterized in that: The steps include: Classify the parts according to their life cycle and order quantity to obtain the categories of the parts; Build a parts labeling system based on parts categories; Pre-process the parts label data according to the parts label system; Construct component demand prediction features based on preprocessed component label data; Constructing a prediction model according to the component demand prediction characteristics; The demand for the components is predicted based on the prediction model.

2. The method for forecasting parts demand according to claim 1, characterized in that: The step of classifying the parts according to the life cycle and order quantity of the parts to obtain the categories of the parts specifically includes: Collect historical spare parts demand data for components; Based on historical spare parts demand data, formulate classification standards for stable spare parts and unstable spare parts; According to the classification standards, spare parts are divided into two categories: stable spare parts and unstable spare parts.

3. The method for forecasting parts demand according to claim 1, characterized in that: The steps of constructing a component labeling system according to the categories of components specifically include: Set label classification criteria; According to the label classification standards, set the correspondence between label categories and label attributes to build a component label system.

4. The method for forecasting parts demand according to claim 1, characterized in that: The step of preprocessing the component label data according to the component label system specifically includes: According to the parts labeling system, clean the parts labeling data; Merge and integrate cleaned component label data from multiple data sources to ensure data consistency and coherence; Data reduction is performed on the merged and integrated component label data.

5. The method for forecasting parts demand according to claim 1, characterized in that: The step of constructing component demand prediction features based on the preprocessed component label data specifically includes: Establish an initial feature set and filter features; Encode categorical features and convert them into numerical features; Scale numerical features to make different features comparable in value; According to business needs, multiple features are combined to form new features.

6. The method for forecasting parts demand according to claim 1, characterized in that: The step of constructing a prediction model according to the component demand prediction characteristics specifically includes: Select a machine learning model based on the characteristics of parts demand data and business needs; Train machine learning models; Validate machine learning models.

7. The method for forecasting parts demand according to any one of claims 1 to 6, characterized in that: The step of predicting the demand for parts according to the prediction model specifically includes: Based on the forecasting model, identifying internal and external factors that affect parts demand; Assess the extent to which internal and external factors affect the demand for parts; Based on the degree of impact, the demand for the components is predicted.

8. A component demand forecasting device, characterized in that: include: A classification module is used to classify parts according to their life cycle and order quantity to obtain the categories of parts; The first building module is used to build a parts labeling system according to the categories of parts; A preprocessing module is used to preprocess the component label data according to the component label system; The second construction module is used to construct component demand prediction features based on the preprocessed component label data; A third building module is used to build a prediction model according to the component demand prediction characteristics; The prediction module is used to predict the demand for the parts according to the prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the component demand forecasting method as described in any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the component demand forecasting method according to any one of claims 1 to 7 are implemented.

Images