Product cost prediction method, system and electronic device of multi-mode interpolation strategy

Abstract

This invention proposes a product cost prediction method, system, and electronic device using a multi-mode interpolation strategy, relating to the field of data processing technology. The method includes acquiring product cost data to be predicted; determining whether missing values ​​exist in the product cost data; if missing values ​​exist, judging the significance of the trend and seasonality of the cost data based on set rules, and dynamically interpolating for each case to obtain the interpolation result corresponding to the missing value; and inserting the interpolation result into the corresponding position in the cost data to be predicted, thereby achieving product cost prediction. This invention performs qualitative analysis of the trend and seasonality of the cost data to be predicted, classifies cases according to the significance of the trend and seasonality, and automatically assigns different interpolation models to form interpolated data. This ensures that the exponential smoothing algorithm remains usable even when the time series is interrupted, and guarantees prediction accuracy.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a product cost prediction method, system and electronic equipment using multi-mode interpolation strategies. Background Technology

[0002] Currently, most companies use time series models for cost forecasting. Seasonal exponential smoothing algorithms are widely used for short-term cost forecasting because they can capture level, trend, and seasonal variations. However, this algorithm has high requirements for the continuity of the input series; if there is an interruption in the period, the model will be unable to proceed with the calculation normally, leading to forecast failure.

[0003] In actual manufacturing, a product may have multiple process routes, resulting in different cost breakdowns. During a certain period, some cost components may not be used, causing a period interruption. Traditional time-series-based exponential smoothing algorithms cannot provide reasonable forecasts at these interruptions. Current solutions include using predictive models to assign values, ignoring the value, or treating it as zero. However, simple interpolation or ignoring these values ​​can easily introduce noise or lose key patterns, affecting the accuracy of cost forecasting. Furthermore, inappropriate interpolation strategies will also negatively impact the accuracy of subsequent cost forecasts. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, this invention provides a product cost forecasting method, system, and electronic device with a multi-mode interpolation strategy. It performs qualitative analysis on the trend and seasonality of the cost data to be predicted, classifies the cases according to the significance of the data trend and seasonality, and automatically assigns different interpolation models to form interpolated data. This ensures that the exponential smoothing algorithm is still usable when the time series is interrupted, and guarantees the prediction accuracy.

[0005] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0006] The first aspect of this invention provides a product cost prediction method using a multi-mode interpolation strategy.

[0007] A product cost forecasting method using a multi-modal interpolation strategy includes the following steps:

[0008] Identify the products whose costs need to be predicted and obtain the predicted cost data for those products.

[0009] Determine if there are missing values ​​in the product's forecast cost data;

[0010] If missing values ​​exist, the trend and seasonality of the cost data to be predicted are calculated;

[0011] Based on the established rules, the significance of the trend and seasonality of the cost data to be predicted is judged, resulting in four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant.

[0012] Based on the four judgment results, the missing values ​​are dynamically interpolated to obtain the interpolation results corresponding to the missing values;

[0013] The interpolation result is inserted into the corresponding position of the cost data to be predicted, thereby achieving product cost prediction.

[0014] A second aspect of the present invention provides a product cost prediction system using a multi-mode interpolation strategy.

[0015] A product cost forecasting system using a multi-modal interpolation strategy includes:

[0016] The data acquisition module is configured to: determine the product whose cost is to be predicted, and acquire the product's cost data to be predicted;

[0017] The missing value identification module is configured to determine whether there are missing values ​​in the product's cost data to be predicted;

[0018] The qualitative calculation module is configured to calculate the trend and seasonality of the cost data to be predicted if missing values ​​exist.

[0019] The qualitative judgment module is configured to: judge the significance of trend and seasonality of the cost data to be predicted based on the set rules, and obtain four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant.

[0020] The dynamic interpolation module is configured to: perform dynamic interpolation on the missing values ​​based on four judgment results, and obtain the interpolation results corresponding to the missing values;

[0021] The prediction module is configured to insert the interpolation results into the corresponding positions of the cost data to be predicted, thereby achieving product cost prediction.

[0022] A third aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the product cost prediction method of the multi-mode interpolation strategy as described in the first aspect of the present invention.

[0023] The above one or more technical solutions have the following beneficial effects:

[0024] This invention discloses a product cost prediction method, system, and electronic device using a multi-mode interpolation strategy. It proposes a fuzzy logic-based method for determining the significance of seasonality and trend, and an adaptive selection mechanism for the multi-mode interpolation strategy. First, qualitative calculations of trend and seasonality are performed on the cost data to be predicted. Then, based on the significance of the qualitative calculation results, four significance levels are identified, and dynamic interpolation calculations are performed for each of the four levels. The interpolated results are then inserted into the corresponding positions in the cost data to be predicted, thus achieving product cost prediction. Based on the results of qualitative analysis, this invention employs different interpolation strategies when the trend and seasonality have different degrees of significance, achieving accurate interpolation calculations.

[0025] This invention proposes an intelligent interpolation mechanism that integrates fuzzy rules and qualitative analysis, which can automatically identify the interruption position of a time series and adaptively select an interpolation strategy based on the significance of trends and seasonality.

[0026] This invention significantly improves the robustness and applicability of the algorithm in discontinuous time series scenarios while ensuring prediction accuracy.

[0027] This invention outputs detailed cost forecasts at the element level, providing more refined decision support for enterprise cost control.

[0028] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0029] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0030] Figure 1 This is a flowchart of the method in Example 1.

[0031] Figure 2 This is a flowchart of the overall method in Embodiment 1.

[0032] Figure 3 This is a line graph of the data when no time series interruption occurred in Example 1.

[0033] Figure 4 This is a line graph of data from Example 1, showing the calculation of missing value interpolation using the full model.

[0034] Figure 5 This is a line graph of data from Example 1, showing the interpolation prediction method used to fill in the previous period for missing value interpolation calculation.

[0035] Figure 6This is a line graph of data used to perform missing value interpolation calculations using the method in Example 1. Detailed Implementation

[0036] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0037] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0038] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0039] Example 1

[0040] Exponential smoothing is a time series model that dynamically weights data from different periods. More recent data has a greater impact, and the weights of each period increase exponentially. It is often used in scenarios involving the prediction of product sales, prices, and production costs.

[0041] However, exponential smoothing algorithms often rely on data over a continuous period, which is often difficult to guarantee in product manufacturing scenarios.

[0042] For ease of understanding, the following description is provided:

[0043] For example, a semi-finished product can be generated first through various raw materials and labor costs, and then product A can be generated through the semi-finished product and machine depreciation costs. At this time, the product cost elements of product A are "raw materials", "labor costs" and "machine depreciation costs".

[0044] During a certain period, due to insufficient production capacity, it may be impossible to produce semi-finished products from multiple raw materials before manufacturing product A. Instead, semi-finished products may be purchased directly and processed by machinery to produce product A. In this case, the product cost elements for product A are "semi-finished products" and "machine depreciation costs." At this time, the elements do not include raw materials, which is equivalent to an interruption in the raw material element.

[0045] In manufacturing product cost forecasting, traditional seasonal exponential smoothing algorithms (such as the Holt-Winters method) heavily rely on continuous time series data. However, in actual production processes, changes in technology, material substitutions, and other factors often cause certain cost elements to be interrupted during specific periods, resulting in discontinuous time series data and leading to the following problems:

[0046] 1. The state parameters (level, trend, seasonality) of the exponential smoothing model cannot be updated normally;

[0047] 2. The predicted results deviate from reality, and the reliability decreases;

[0048] 3. Existing interpolation methods (such as ignoring, zeroing, or simple padding) introduce noise and amplify prediction errors.

[0049] Based on this, this invention proposes an interpolation algorithm based on qualitative analysis, which adopts different interpolation strategies for missing data values ​​when the trend and seasonality of the data are different.

[0050] This embodiment discloses a product cost forecasting method with a multi-mode interpolation strategy. In the process of using the seasonal exponential smoothing algorithm for each factor, the cost data to be predicted is first qualitatively analyzed. The qualitative analysis results are classified according to the significant characteristics of the data, and different interpolation models are automatically assigned to form interpolated data. This ensures that the exponential smoothing algorithm is still usable when it is interrupted during the period, and can continuously output product cost factor-level details with high reliability. This solves the technical problem that the exponential smoothing model cannot be used due to the interruption of some factors during the period.

[0051] When using the seasonal exponential smoothing algorithm to make short-term forecasts of product cost details, a trend coefficient and a seasonality coefficient are defined for iterative updates. The coefficient iteration formula is as follows (based on a multiplicative model):

[0052] (Equation 1)

[0053] (Equation 2)

[0054] (Equation 3)

[0055] (Equation 4)

[0056] in, Represents the level term during period T; This represents the trend term during period T; The seasonal term at time T; This represents the predicted value during the T+K period; , , These represent the horizontal smoothing coefficient, the trend smoothing coefficient, and the seasonal smoothing coefficient, respectively. This represents the actual value during period T; This represents the level term during period T-1; This represents the trend term during period T-1; This indicates the seasonal term in the same season of the previous period; m represents the length of the season. This indicates the seasonal item corresponding to the same season in the previous period of T+K; This indicates the Kth period to be predicted.

[0057] The forecasting process can be defined as an ω-automaton. Assuming that it only forecasts for the next period, it can be formally represented as a six-tuple. Where Q is the set of states, It is the initial state. It is an input symbol. It is the output symbol. It is a state transition function. It is the output function.

[0058] and and satisfy:

[0059] ,express:

[0060] Output function: Given an input symbol in Q state Output a corresponding output symbol .

[0061] ,express:

[0062] State transition function: In Q state, based on the input The status has changed to ,at the same time and They all belong to the state set Q.

[0063] In this embodiment, Equation 1 represents the iterative formula for the horizontal component, Equation 2 represents the iterative formula for the trend component, and Equation 3 represents the iterative formula for the seasonal component. Equations 1, 2, and 3 constitute the state transition function. Equation 4 constitutes the output function. .

[0064] Q consists of an infinite number of triples. ( )constitute. Let T represent the triplet during time T, which is the state at time T.

[0065] For each input value , This represents the input value during period T. The structure is ( , ), ( , ) represents a time-price pair, where Indicates time, express The actual value during the period, i.e. The actual price during the period. (Yes) And the automaton outputs a corresponding result. , express The predicted value during the period.

[0066] Under this definition, assume an infinitely long, non-repeating input sequence. If it exists The automaton will then enter an unusable state. Among these, This represents a time-price pair during period 1. This represents the time-price pair during period 2. This represents the time-price pair during period i; This represents time-price pairs arranged in chronological order.

[0067] The method of this embodiment will now be explained in detail.

[0068] like Figure 1 As shown, the product cost forecasting method using a multi-mode interpolation strategy includes the following steps:

[0069] Identify the products whose costs need to be predicted and obtain the predicted cost data for those products.

[0070] Determine if there are missing values ​​in the product's forecast cost data;

[0071] If missing values ​​exist, the trend and seasonality of the cost data to be predicted are calculated;

[0072] Based on the established rules, the significance of the trend and seasonality of the cost data to be predicted is judged, resulting in four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant.

[0073] Based on the four judgment results, the missing values ​​are dynamically interpolated to obtain the interpolation results corresponding to the missing values;

[0074] The interpolation result is inserted into the corresponding position of the cost data to be predicted, thereby achieving product cost prediction.

[0075] Furthermore, in this embodiment, the specific process for obtaining the product's cost data to be predicted is as follows:

[0076] Multiple process routes for the product whose cost is to be predicted are identified, and then the cost elements that constitute the cost of the product for the corresponding process route are determined.

[0077] Obtain detailed historical cost data for each cost element, including product, period, production workshop, element number, and the amount of the corresponding period cost element;

[0078] The historical cost data of different cost elements are arranged by period, grouped according to the product cost structure to which the cost elements belong, and the amounts of the same period, the same product, and the same element are weighted and averaged to obtain the cost data to be predicted.

[0079] A process route refers to different processes used to produce the same product. In this embodiment, different cost structures for the same product are identified as different processes.

[0080] The reason why there are multiple process routes for a product whose cost needs to be predicted is that the product cost may be composed of different cost elements at different stages of production. Different cost elements form different cost structures, and different cost structures correspond to different process routes.

[0081] Grouping historical cost data for multiple elements according to the above dimensions and then weighting them separately can be understood as follows: the cost structure of the same batch of products produced in the same workshop during a certain period is exactly the same.

[0082] Furthermore, it is necessary to determine whether there are missing values ​​in the product's cost data to be predicted, specifically including:

[0083] In the cost data to be predicted, determine whether there is an interruption in the time series of adjacent cost data. If there is an interruption, there are missing values; otherwise, there are no missing values.

[0084] Furthermore, in this embodiment, when performing qualitative calculations on the product's to-be-predicted cost data:

[0085] The trend of the cost data to be predicted is calculated based on the numerical value of the cost data to be predicted, and the seasonality of the cost data to be predicted is calculated by setting a seasonal significant membership function based on fuzzy rules.

[0086] More specifically, calculate the trend significance index to measure the trend of the forecast cost data:

[0087] (Equation 5)

[0088] in, The trend significance index represents time i; This represents the trend term during period i; This represents the predicted value for period i; This represents the trend term during period 1; This represents the actual value during period 1; Represents the trend term over period n; This represents the actual value during period n.

[0089] Based on the trend significance membership function, the significance of the trend in the cost data to be predicted is determined. The specific method is as follows:

[0090] If the significant membership degree of the trend calculated by the significant membership function is 0, it means that the trend is not significant.

[0091] If the trend significance membership degree calculated by the trend significance membership function is 1, then the trend is significant.

[0092] The significant trend membership function is as follows:

[0093] (Equation 6)

[0094] in, and The set boundary value.

[0095] More specifically, based on fuzzy rules, a seasonal significant membership function is set to calculate the seasonality of the cost data to be predicted, including:

[0096] Based on the seasonality coefficient in the seasonality index smoothing algorithm, calculate the seasonality parameters of the cost data to be predicted.

[0097] Set multiple boundary values;

[0098] Based on the calculated seasonal parameters of the cost data to be predicted and the magnitude of multiple threshold values, determine the threshold value range in which the seasonal parameters of the cost data to be predicted fall.

[0099] Based on different threshold ranges, different seasonal significant membership functions are set to calculate seasonal significant membership.

[0100] The seasonality of the cost data to be predicted is measured by seasonal significant membership.

[0101] Trends can be judged directly based on numerical values, while the strength of seasonality needs to be determined by clarifying the fuzzy function and calculating the seasonal significance index. :

[0102] (Equation 7)

[0103] in, Indicates the seasonal significance index; This represents the seasonal term during period i; Indicates average; This indicates the seasonal items during period 1; This represents the seasonal term during period n.

[0104] The membership parameter based on fuzzy logic allows different delimitation values ​​to be set for different domains, such as a, b, c, and d, thereby determining the seasonal correlation fuzzy rule in Equation 8. The specific delimitation values ​​can be evaluated and set by domain experts from different fields.

[0105] The specific seasonal significant membership function is expressed as follows:

[0106] (Equation 8)

[0107] Where a, b, c, and d are all delimited values, and a b c d; Significant membership degree due to seasonality; This indicates the seasonal significance index.

[0108] parameter As a basis for determining whether the current data has obvious seasonality, it is integrated into the calculation process of dynamic interpolation.

[0109] Furthermore, based on the calculated seasonal significance membership value, the degree of seasonality of the cost data to be predicted is determined. When the seasonal significance membership value is 1, it indicates that the cost data is seasonally significant; when the seasonal significance membership value is 0, it indicates that the seasonality is not significant. Within the range of 0-1, the closer to 1, the more significant the seasonality, and the closer to 0, the less significant the seasonality.

[0110] When qualitatively modeling the significance of trends and seasonality, trends and seasonality are categorized according to their significance as follows:

[0111]

[0112] ;

[0113] in, Indicates a trend. Indicates seasonality.

[0114] When trend or seasonal characteristics are not significant, the trend or seasonality coefficients are often caused by noise. If the model is used directly for prediction, the noise will be amplified. Therefore, this embodiment realizes dynamic interpolation calculation based on the qualitative analysis of whether the corresponding factors are significant.

[0115] Furthermore, based on the four judgment results, dynamic interpolation is performed on the missing values ​​to obtain the interpolation results corresponding to the missing values. The specific dynamic interpolation formula is as follows:

[0116] (Equation 9)

[0117] Equation 9's implementation depends on the generation of fuzzy variables Trend and Seasonal. Among them, This represents the predicted value during period i+1; This represents the actual value during period i; This represents the trend term during period i; Represents the level term during period i; Indicates seasonal significance of membership; This represents the seasonal term corresponding to the previous period of the same season during period i+1.

[0118] In summary, the embodiments of the present invention, by improving the exponential smoothing algorithm and using different interpolation algorithms through qualitative analysis, resolve missing... Improve the system to prevent the automaton from becoming unusable.

[0119] The intelligent interpolation algorithm in Equation 9 can effectively extract the most significant features (trend, seasonality, both or none) of the current cost data to be predicted, effectively reducing the noise introduced by unreasonable interpolation algorithms.

[0120] More specifically, as a concrete example, the implementation scheme is as follows:

[0121] 1. Data model preparation and initialization:

[0122] For each selected product, a weighted average of the costs of various elements from previous calculation sheets is used to form an input sequence. ,according to The initial state is .in, This represents a time-price pair during period 1. This represents the time-price pair during period 2. Represents time-price pairs over period n; This represents time-price pairs arranged in chronological order.

[0123] Then output a As the next period (i.e. The predicted value during the period.

[0124] 2. Interrupt detection and interpolation triggering:

[0125] During the operation of the cost prediction system, It is arranged strictly in chronological order. If it appears In this intermittent situation, the system recognizes that an interruption has occurred. It is determined that interpolation processing is required.

[0126] in, This represents the time-price pair during period i; This represents the time-price pair during period i+2; exist If it does not appear between i+1, it means that an interruption occurred during i+1.

[0127] 3. Qualitative analysis of the significance of trends and seasonality:

[0128] use Perform calculations. This represents the state of the automaton during period i. The fuzzy variable Trend is refined based on a specified trend threshold; subsequently, the fuzzy variable Seasonal is refined based on the membership function.

[0129] 4. Multi-mode intelligent interpolation:

[0130] Finally, the interpolation was calculated according to Equation 6. ,Will insert The missing segment in the automaton allows it to return to a steady state. Among them, This represents the predicted value during period i+1.

[0131] Next, the accompanying drawings in this embodiment will be explained in detail:

[0132] like Figure 2 As shown, when executing the scheme of this embodiment, it is first necessary to obtain the initialization model. In this embodiment, the initialization model refers to the exponential smoothing algorithm model. The time series and the cost data to be predicted are input, and the continuity of the time series of the data in the cost data to be predicted is judged. If the time series is continuous, it means that there are no missing values. The prediction result is directly output and the process ends.

[0133] If not, it means that there are missing values ​​in the cost data to be predicted. The specific implementation steps of this plan need to be executed to perform qualitative calculations on the cost data to be predicted, namely trend and seasonality calculations, and to make a significance judgment.

[0134] When making a significance judgment, it is necessary to calculate the seasonal significance membership function and perform dynamic interpolation calculation based on fuzzy rules to obtain the interpolation result;

[0135] The interpolation results are inserted into the corresponding positions of the cost data to be predicted, thereby achieving product cost prediction and improving the time series of the cost data to be predicted that originally had missing values.

[0136] The improved time series and data are then used as new input data for the next round of iterative judgment until all missing values ​​are filled in, at which point the entire execution process ends.

[0137] Figure 3 This represents the scenario where no time series interruption occurs. Figure 3 Based on this, predictions were made after manually deleting the time series and numerical values ​​for periods 3 and 7, and then using different interpolation methods. Specifically, Figure 4 The results are from missing value interpolation using the full model. Figure 5 This represents the results of missing value calculation using the interpolation prediction method filled in the previous period. Figure 6 The result is the interpolation calculation performed using the method of this embodiment.

[0138] exist Figure 3 In the graph, the horizontal axis represents the period, and the vertical axis represents the predicted value. The blue solid line represents the actual value, and the orange solid line represents the predicted value.

[0139] Figure 4 In the graph, the horizontal axis represents the period, and the vertical axis represents the predicted value. Specifically, the blue solid line represents the actual value, the orange solid line represents the predicted value, the yellow dashed line represents the interpolation of the actual value by the full model output, and the gray dashed line represents the predicted value based on the interpolation output.

[0140] Figure 5 In the graph, the horizontal axis represents the period, and the vertical axis represents the predicted value. Specifically, the blue solid line represents the actual value, the orange solid line represents the predicted value, the yellow dashed line represents the interpolation of the actual value using the interpolation prediction method filled with the previous period, and the gray dashed line represents the predicted value based on the interpolation output.

[0141] Figure 6 In the diagram, the horizontal axis represents the period, and the vertical axis represents the predicted value. The blue solid line represents the actual value, the orange solid line represents the predicted value, the yellow dashed line represents the interpolation of the actual value using the method of this embodiment, and the gray dashed line represents the predicted value based on the interpolation.

[0142] Figure 4 , Figure 5 and Figure 6 To initialize the model using two complete cycles when α, β, and γ are specified as 0.25 in the iterative formulas for the coefficients of the exponential smoothing model (i.e., Equations 1, 2, 3, and 4).

[0143] Figure 6 To establish the seasonal significance parameters a, b, c, and d, they were set to 0.25, 0.8, 1.2, and 1.5 respectively. and The prediction results when the values ​​are set to 0.9 and 1.1 respectively.

[0144] by Figure 3 Based on the premise of no data interruption during the period, the following aspects are considered:

[0145] 1. Since the prediction model requires data from two periods to initialize the model and a third period for short-term prediction, the predicted value after interpolation should retain curve characteristics similar to the prediction before interpolation as much as possible to prevent the loss of features (seasonal features, trend features).

[0146] 2. After interpolation, the difference between the predicted value and the subsequent actual value should be as small as possible, indicating that the prediction model after interpolation can still extract the trend and seasonal characteristics.

[0147] By comparison Figure 4 and Figure 6 It can be observed that, due to the heavy weighting of recent data in the exponential smoothing algorithm, interpolation using the full prediction model leads to the loss of the periodicity and trend characteristics of the curve, especially during the 8-12 period (orange line), and the difference in predictions increases in the subsequent short period.

[0148] contrast Figure 5 and Figure 6 It can be observed that although the difference in prediction results is small when using the previous period's value for filling, the characteristics of the curve are lost, especially in the period 8-12 (orange line).

[0149] By comparison Figure 3 and Figure 6 It was found that when using the intelligent interpolation algorithm proposed in this embodiment, the influence of features and noise can be distinguished as much as possible, preserving curve features to the greatest extent while ensuring the accuracy of prediction results in the short term.

[0150] The intelligent interpolation mechanism proposed in this invention, which integrates fuzzy rules and qualitative analysis, can automatically identify the interruption position of a time series and adaptively select an interpolation strategy based on the significance of trends and seasonality. While ensuring prediction accuracy, it significantly improves the robustness and applicability of the algorithm in discontinuous time series scenarios.

[0151] Example 2

[0152] This embodiment discloses a product cost prediction system using a multi-mode interpolation strategy.

[0153] A product cost forecasting system using a multi-modal interpolation strategy includes:

[0154] The data acquisition module is configured to: determine the product whose cost is to be predicted, and acquire the product's cost data to be predicted;

[0155] The missing value identification module is configured to determine whether there are missing values ​​in the product's cost data to be predicted;

[0156] The qualitative calculation module is configured to calculate the trend and seasonality of the cost data to be predicted if missing values ​​exist.

[0157] The qualitative judgment module is configured to: judge the significance of trend and seasonality of the cost data to be predicted based on the set rules, and obtain four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant.

[0158] The dynamic interpolation module is configured to: perform dynamic interpolation on the missing values ​​based on four judgment results, and obtain the interpolation results corresponding to the missing values;

[0159] The prediction module is configured to insert the interpolation results into the corresponding positions of the cost data to be predicted, thereby achieving product cost prediction.

[0160] Example 3

[0161] The purpose of this embodiment is to provide an electronic device.

[0162] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the product cost prediction method using a multi-mode interpolation strategy as described in Embodiment 1 of this disclosure.

[0163] The steps and methods involved in the apparatuses of Embodiments 2 and 3 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0164] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0165] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for product cost prediction using a multi-mode interpolation strategy, characterized in that, Includes the following steps: Identify the products whose costs need to be predicted and obtain the predicted cost data for those products. Determine if there are missing values ​​in the product's forecast cost data; If missing values ​​exist, the trend and seasonality of the cost data to be predicted are calculated; Based on the established rules, the significance of the trend and seasonality of the cost data to be predicted is judged, resulting in four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant. Based on the four judgment results, the missing values ​​are dynamically interpolated to obtain the interpolation results corresponding to the missing values; The interpolation result is inserted into the corresponding position of the cost data to be predicted to achieve product cost prediction. The trend of the cost data to be predicted is calculated based on the magnitude of the data, and the seasonality of the cost data to be predicted is calculated by setting a seasonal significant membership function based on fuzzy rules. The seasonality parameter of the cost data to be predicted is calculated using the following formula: ； wherein, denotes the seasonality index; denotes the seasonality term for period i; denotes the average; denotes the seasonality term for period 1; denotes the seasonality term for period n; The seasonal significant membership function is calculated using the following formula: ； wherein a, b, c, d are all limit values, wherein a b c d; is the seasonal significant membership; denotes the seasonal significant index; Based on the calculated seasonal significance membership values, the degree of seasonal significance of the cost data to be predicted is determined, including: When the seasonal significance membership degree is 1, it indicates that the seasonality is significant. When the seasonal significance membership is 0, it indicates that the seasonality is not significant. When the seasonal significance membership degree is in the range of 0-1, the difference from 1 within the first set range indicates that the seasonality is more significant, and the difference from 0 within the second set range indicates that the seasonality is less significant. Based on the four judgment results, dynamic interpolation is performed on the missing values ​​to obtain the interpolation results corresponding to the missing values. The dynamic interpolation formula is as follows: ； wherein, represents the predicted value during i+1; represents the actual value during i; represents the trend item during i; represents the level item during i; represents the seasonality significant membership degree; represents the seasonal item during i+1 in the same season of the last period; based on the numerical size of the cost data to be predicted, the trend of the predicted cost data is measured, specifically: ； in, The index representing the significance of the trend at time i; This represents the trend term during period i; This represents the predicted value for period i; This represents the trend term during period 1; This represents the actual value during period 1; Represents the trend term over period n; Represents the actual value during period n; Based on the trend significance membership function, the significance of the trend in the cost data to be predicted is determined. The specific method is as follows: If the significant membership degree of the trend calculated by the significant membership function is 0, it means that the trend is not significant. If the trend significance membership degree calculated by the trend significance membership function is 1, then the trend is significant. The significant membership function of the trend is specifically: ； in, and The set boundary value.

2. The product cost prediction method using a multi-mode interpolation strategy as described in claim 1, characterized in that, The specific process for obtaining the predicted cost data for the product is as follows: Multiple process routes for the product whose cost is to be predicted are identified, and then the cost elements of the product cost for the corresponding process route are determined. Obtain detailed historical cost data for each cost element, including product, period, production workshop, element number, and the amount of the corresponding period cost element; The historical cost data of different cost elements are arranged by period, grouped according to the product cost structure to which the cost elements belong, and the amounts of the same period, the same product, and the same element are weighted and averaged to obtain the cost data to be predicted.

3. The product cost prediction method using a multi-mode interpolation strategy as described in claim 2, characterized in that, Determining whether there are missing values ​​in the product's forecast cost data specifically includes: In the cost data to be predicted, determine whether there is an interruption in the time series of adjacent cost data. If there is an interruption, there are missing values; otherwise, there are no missing values.

4. The product cost prediction method using a multi-mode interpolation strategy as described in claim 1, characterized in that, Based on fuzzy rules, a seasonal significant membership function is set to calculate the seasonality of the cost data to be predicted, specifically including: Based on the seasonality coefficient in the seasonality index smoothing algorithm, calculate the seasonality parameters of the cost data to be predicted. Set multiple threshold values; Based on the calculated seasonal parameters of the cost data to be predicted and the magnitude of multiple threshold values, determine the threshold value range in which the seasonal parameters of the cost data to be predicted fall. Based on different threshold ranges, different seasonal significant membership functions are set to calculate seasonal significant membership. The seasonality of the cost data to be predicted is measured by seasonal significant membership.

5. A product cost prediction system based on a multi-mode interpolation strategy as described in any one of claims 1-4, characterized in that, include: The data acquisition module is configured to: determine the product whose cost is to be predicted, and acquire the product's cost data to be predicted; The missing value identification module is configured to determine whether there are missing values ​​in the product's cost data to be predicted; The qualitative calculation module is configured to calculate the trend and seasonality of the cost data to be predicted if missing values ​​exist. The qualitative judgment module is configured to: judge the significance of trend and seasonality of the cost data to be predicted based on the set rules, and obtain four judgment results: both seasonality and trend are significant, seasonality is significant but trend is not significant, seasonality is not significant but trend is significant, and neither seasonality nor trend is significant. The dynamic interpolation module is configured to: perform dynamic interpolation on the missing values ​​based on four judgment results, and obtain the interpolation results corresponding to the missing values; The prediction module is configured to insert the interpolation results into the corresponding positions of the cost data to be predicted, thereby achieving product cost prediction.

6. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the product cost prediction method of the multi-mode interpolation strategy as described in any one of claims 1-4.

Images