A method for optimizing enterprise innovation input based on DEA prediction and evaluation

Abstract

The present invention discloses a method for optimizing enterprise innovation input based on DEA forecasting evaluation, which relates to the technical field of enterprise innovation efficiency evaluation, including determining the input index, output index and environmental index of an enterprise, collecting and preprocessing the data of relevant enterprises; establishing a forecasting model for corresponding indicators, and determining the input variables through the Granger causality detection method, and performing model training; based on the DEA-BCC model, evaluating the innovation efficiency of the current year, and adjusting the innovation input of the next year according to the evaluation results; then obtaining the output index of the next year according to the optimized innovation input; according to the innovation input and output index of the enterprise in the next year, using the DEA-BCC model for cyclic calculation, obtaining the input efficiency score and slack variables of the enterprise in the next year, and forming an investment plan according to the calculation results. Therefore, the above-mentioned method for optimizing enterprise innovation input based on DEA forecasting evaluation can guide enterprises to optimize their internal resources and improve innovation efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise innovation efficiency evaluation, and in particular to an enterprise innovation input optimization method based on DEA prediction evaluation. Background Art

[0002] Today's society has entered the era of knowledge economy. Innovation is an important driving force for promoting economic development and is closely related to national prosperity, national prosperity, and people's happiness. State-owned enterprises are the boosters of the economy. Accurate evaluation of the innovation efficiency of state-owned enterprises' scientific and technological innovation will help state-owned enterprises tap their innovation potential and improve their innovation capabilities, which will have a far-reaching impact on national economic development and international competitiveness. DEA is an important method for evaluating innovation efficiency. Enterprises can optimize the allocation of internal innovation inputs based on the DEA evaluation results to improve technological innovation performance.

[0003] At present, there are the following problems when using innovation efficiency evaluation to optimize innovation resources: the evaluation is based on facts that have already occurred, and the use of evaluation results to configure innovation inputs is effective for outputs that have already occurred, that is, when the output remains unchanged, the conclusion is accurate. However, future outputs will change due to changes in inputs and changes in the environment. The conclusions obtained in this way cannot be applied to future scenarios. In essence, the purpose of optimizing resource inputs is also to improve future innovation efficiency, but there is currently no relevant evaluation and decision-making method.

[0004] Based on this, this patent proposes a predictive DEA evaluation model that can perform DEA evaluation on the future innovation efficiency of enterprises and be used to guide enterprises to optimize their innovation investment. Summary of the invention

[0005] The purpose of this invention is to provide an enterprise innovation input optimization method based on DEA forecasting and evaluation, which can predict and evaluate future input-output, and then guide enterprises to optimize their internal resources to improve innovation efficiency.

[0006] To achieve the above object, the present invention provides an enterprise innovation investment optimization method based on DEA prediction evaluation, comprising the following steps: S1. Determine the input indicators, output indicators and environmental indicators of the enterprise, collect and pre-process the data of relevant enterprises; S2. Determine the Granger cause of each output indicator through the Granger causality detection method, and use it as the input variable of the corresponding prediction model for model training; S3. Based on the DEA-BCC model, analyze the target enterprises and selected enterprises, obtain the efficiency score of each decision-making unit, and adjust the innovation investment for the next year according to the evaluation results; S4. Based on the innovation input for the next year, use the trained prediction model to obtain the output indicators for the next year; S5. Based on the enterprise's innovation input and output indicators for the next year, use the DEA-BCC model to perform cyclic calculations to obtain the enterprise's input efficiency score and slack variables for the next year, and form an investment plan based on the calculation results.

[0007] Preferably, input indicators include total funding input, innovation funding input, number of R&D personnel and number of advanced equipment; output indicators include number of patent applications, number of approved patents and operating income; environmental indicators include market prosperity index and Herfindahl index.

[0008] Preferably, the prediction model includes an operating income prediction model, a patent application prediction model and a patent authorization quantity prediction model.

[0009] Preferably, the Granger causes of each output indicator are determined, including: First, the VAR model is established using various output indicators, the size of the sliding window is determined, and the least squares principle is used to solve and calculate the various parameters of the VAR model; Secondly, the VAR model is established using input indicators and output indicators, and the F test is conducted on each input indicator through the significance level to obtain the Granger cause of the corresponding output indicator; Then, a multivariate prediction model is established, the Granger causes of the corresponding output indicators are used as input, and the corresponding model is used for training until the set standards are met, completing the multivariate prediction model training.

[0010] Preferably, the patent grant quantity prediction model needs to detect the Granger causality between the patent application quantity and the patent grant quantity.

[0011] Preferably, step S3 includes: S31. Take the enterprise as the decision-making unit, select the input-oriented model, and input the input indicators and output indicators of the current year into the DEA-BCC model; S32. Solve the DEA-BCC model, obtain the efficiency score of the decision-making unit, and evaluate the technical efficiency based on the calculated input efficiency score and slack variables.

[0012] Preferably, the DEA-BCC model is as follows: ; In the formula, and Respectively represent the The decision-making unit Input, outputs, is the index of the decision unit, For the The weight of a decision-making unit, The efficiency score of the input, is the error term, and is the slack variable, where It indicates the amount of output that can be increased by the decision-making unit under the current input level. It indicates the amount of input that can be reduced under current production conditions without affecting the output level.

[0013] Therefore, the present invention adopts the above-mentioned enterprise innovation investment optimization method based on DEA prediction and evaluation, which has the following technical effects: By establishing a model to predict future output, the DEA model can be used to evaluate current innovation efficiency, optimize innovation inputs based on the evaluation results, and iterate calculations until the innovation efficiency meets the requirements, thereby guiding enterprises to optimize their internal resources and improve innovation efficiency.

[0014] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the enterprise innovation input optimization method based on DEA forecasting and evaluation. DETAILED DESCRIPTION

[0016] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0017] like Figure 1 As shown, the present invention provides an enterprise innovation investment optimization method based on DEA prediction evaluation, comprising the following steps: S1. Select 20 enterprises with similar scale, type and establishment years as the enterprise to be evaluated, establish input index X, output index C and environmental index N, and collect and pre-process the data.

[0018] Among them, the input indicators of enterprises include: total funding input of enterprises J1, innovation funding input of enterprises J2, number of R&D personnel F, number of advanced equipment E, etc. Output indicators include: number of patent applications Z1, number of patents approved Z2, operating income Y, etc. Environmental parameters include: market prosperity index M, Herfindahl index H, etc. The data of each indicator are collected and counted on an annual basis. According to the shortest establishment period of the selected enterprises, data from the second year of establishment to the last year are obtained. The detailed description is shown in Table 1.

[0019] Table 1 Description of each indicator ;

[0020] S2. The Granger causality detection method is used to detect the Granger causes of various output indicators, and they are used as inputs of various prediction models for model training.

[0021] 1. Train a prediction model for operating income Y to predict the operating income of the target company, including the following steps: (1) Determine variables: Use the Granger causality detection method to test each input indicator to determine which indicators are the Granger causes of operating income Y, and use these indicators as input variables of the prediction model, as follows: First, use the operating income Y to build a VAR model, and the size of the sliding window is selected as , we get the constrained residual sum of squares , the expression is: ; ; In the formula, is the predicted value; is the actual value; is a constant term, is the lag term coefficient, ; is a random influencing factor. By using the least squares method, the residual sum of squares is minimized and the various parameters of the VAR equation can be calculated.

[0022] Then, each input indicator X and operating income Y are used to build a VAR model to obtain the constrained residual sum of squares. , the expression is: ; ; In the formula, is the predicted value after adding the input indicator X; is a constant term; , is the lag term coefficient, ; is a random influencing factor.

[0023] Finally, at the significance level Next, an F test is performed on each input indicator, as follows: ; if , this input indicator is the Granger cause of operating income. Based on this, it is possible to detect which indicators are the Granger causes of operating income.

[0024] (2) Establish a multivariate prediction model: Use the SVM model or multivariate linear regression model. The input of the prediction model is the operating income and its Granger cause, and the output is the operating income. If there is no Granger cause, the input is only the operating income.

[0025] (3) Input the operating income and its Granger causes into the SVM model or multivariate linear regression model for training. Select the sliding window size that corresponds to the maximum F value. Use 80% of the data for training and the remaining 20% ​​of the data for testing. When the test accuracy reaches the pre-set standard, the multivariate prediction model training is completed.

[0026] 2. Similarly, train a patent application prediction model to predict the number of patent applications of the target enterprise. Specifically, first find the Granger cause of the patent application quantity Z1, and then use these indicators and the patent application quantity Z1 as input to predict the patent application quantity Z1. Here, the prediction model also uses SVM or multivariate linear regression.

[0027] 3. Train a patent authorization quantity Z2 prediction model to predict the number of authorization applications of the target enterprise. Since patent authorization has a certain lag, it is necessary to detect whether the patent application quantity Z1 is the Granger cause of Z2. If so, Z1 should also be used as the input of the patent authorization prediction model.

[0028] S3. Conduct DEA analysis on the target enterprise and the selected enterprises, as follows: S31. Take the enterprise as a decision-making unit (DMU), select the input-oriented model, and bring the input indicators and output indicators of the most recent year into the DEA-BCC model, as follows: ;

[0029] in, and Respectively represent the The decision-making unit Input, Outputs; is the index of the decision-making unit; For the The weight of each decision-making unit; is the input efficiency score; is the error term, with a value of 10 -6 ; and is the slack variable, where It indicates the amount of output that can be increased by the decision-making unit under the current input level. It indicates the amount of input that can be reduced under current production conditions without affecting the output level.

[0030] S32, solve the DEA model, use DEA software to execute the DEA model, calculate the efficiency score of each DMU, ​​and , and Evaluate technical efficiency as follows: For the target enterprise, , , is optimal; when , or Becoming weakly effective can reduce the input Or increase output ; DEA is invalid. The closer to 1, the more efficient it is. If the original output remains unchanged, the input will be reduced to Ratio, can be , to achieve optimal efficiency.

[0031] S33. Enterprises shall adjust their innovation investment for the next year based on the evaluation results.

[0032] S4. Predict the operating income, patent applications and patent authorizations for the next year. Specifically, the adjusted innovation input is used as input, and the corresponding models are used to predict the operating income, patent applications and patent authorizations for the next year.

[0033] S5. Bring the adjusted innovation input and the predicted operating income, patent application volume and patent authorization volume of the next year into the DEA-BCC model to calculate , and If DEA is valid, then the calculation is exited and the resulting plan is invested. Otherwise, steps S4-S6 are repeated until the loop ends. Here, the number of loops is set to 20, as follows: For enterprises that are initially evaluated as weak DEA effective, , or (Not 0) When one of the items reaches the minimum and does not change for two consecutive cycles, the cycle ends and the final result is output as the investment benchmark for the next year.

[0034] For enterprises that are initially evaluated as DEA invalid, When it reaches the maximum and does not change for two consecutive cycles, the cycle ends and the final result is output as the investment benchmark for the next year.

[0035] If the cycle exceeds 20 times and there is still no stability, select The maximum output result will be used as the investment benchmark for the next year.

[0036] Therefore, the present invention adopts the above-mentioned enterprise innovation input optimization method based on DEA prediction and evaluation, which can guide enterprises to optimize innovation input and achieve improved innovation efficiency in the future.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for optimizing enterprise innovation input based on DEA prediction and evaluation, characterized in that: The following steps are involved: S1. Determine the input indicators, output indicators and environmental indicators of the enterprise, collect and pre-process the data of relevant enterprises; S2. Determine the Granger cause of each output indicator through the Granger causality detection method, and use it as the input variable of the corresponding prediction model for model training; S3. Based on the DEA-BCC model, analyze the target enterprises and selected enterprises, obtain the efficiency score of each decision-making unit, and adjust the innovation investment for the next year according to the evaluation results; S4. Based on the innovation input for the next year, use the trained prediction model to obtain the output indicators for the next year; S5. Based on the enterprise's innovation input and output indicators for the next year, use the DEA-BCC model to perform cyclic calculations to obtain the enterprise's input efficiency score and slack variables for the next year, and form an investment plan based on the calculation results.

2. According to claim 1, a method for optimizing enterprise innovation input based on DEA forecasting evaluation is characterized in that: Input indicators include total funding input, innovation funding input, number of R&D personnel and number of advanced equipment; output indicators include the number of patent applications, number of approved patents and operating income; environmental indicators include market prosperity index and Herfindahl index.

3. The enterprise innovation input optimization method based on DEA forecasting evaluation according to claim 1 is characterized in that: The forecasting models include operating income forecasting model, patent application forecasting model and patent authorization quantity forecasting model.

4. The enterprise innovation input optimization method based on DEA forecasting evaluation according to claim 1 is characterized in that: Determine the Granger causes for each output indicator, including: First, the VAR model is established using various output indicators, the size of the sliding window is determined, and the least squares principle is used to solve and calculate the various parameters of the VAR model; Secondly, the VAR model is established using input indicators and output indicators, and the F test is conducted on each input indicator through the significance level to obtain the Granger cause of the corresponding output indicator; Then, a multivariate prediction model is established, the Granger causes of the corresponding output indicators are used as input, and the corresponding model is used for training until the set standards are met, completing the multivariate prediction model training.

5. The enterprise innovation input optimization method based on DEA forecasting evaluation according to claim 3 is characterized in that: The patent authorization quantity prediction model needs to detect the Granger causality between the patent application quantity and the patent authorization quantity.

6. The enterprise innovation input optimization method based on DEA forecasting evaluation according to claim 1 is characterized in that: Step S3 includes: S31. Take the enterprise as the decision-making unit, select the input-oriented model, and input the input indicators and output indicators of the current year into the DEA-BCC model; S32. Solve the DEA-BCC model, obtain the efficiency score of the decision-making unit, and evaluate the technical efficiency based on the calculated input efficiency score and slack variables.

7. The enterprise innovation input optimization method based on DEA forecasting evaluation according to claim 1 is characterized in that: The DEA-BCC model is as follows: ； In the formula, and Respectively represent the The decision-making unit Input, outputs, is the index of the decision unit, For the The weight of a decision-making unit, is the efficiency score of the input, is the error term, and is the slack variable, where It indicates the amount of output that can be increased by the decision-making unit under the current input level. It indicates the amount of input that can be reduced under current production conditions without affecting the output level.

Images