CN106202310A - A kind of method setting up data mining automatic feedback system - Google Patents

Abstract

The present invention provides a kind of method setting up data mining automatic feedback system, for the problem solving to need continuous manual feedback tuning during current data mining realizes.The data mining automatic feedback system that the method is set up, including data segmentation module, evaluation of result module, parameter adjustment module；Three big module cooperative work form feedback, automatically adjust, optimize the parameter of data mining algorithm, carry out data mining than ever and more save human cost.By the automatic adjustment algorithm of parameter efficiently, automatic adjustment algorithm parameter accurately, effective improve parameter adjustment efficiency, reduce the actual waste that traversal parameter value scope too much in automation process causes.

Description

A Method of Establishing Data Mining Automatic Feedback System

technical field

The invention relates to the field of data mining, in particular to a method for establishing an automatic feedback system for data mining.

Background technique

With the rapid development of big data technology, data mining technology is more widely used. Universities, research institutes, governments and technology companies are widely using data mining technology.

A complete data mining process often includes: data preprocessing, data mining algorithm execution, and data result reporting. The most critical step is to obtain the results of data mining through the execution of data mining algorithms. This step often requires a lot of manual intervention and feedback. Manual intervention and feedback are manifested as observing the results of the algorithm execution through the expert's empirical model, and then readjusting the algorithm parameters according to the results to re-execute the algorithm to obtain new results until the data mining obtains satisfactory results. This process often consumes a lot of labor costs, wasting time and energy. Although most data mining algorithms can iterate and converge continuously, the data mining results caused by the initial parameters and local optimal solutions in the calculation process are not ideal, and cannot be solved by the algorithm itself iteratively.

Contents of the invention

The purpose of the present invention is to solve the problem of continuous manual feedback and optimization in the current data mining implementation process, and provide a method for establishing an efficient, realizable, and automatic data mining automatic feedback system.

The present invention relates to a method for establishing an automatic feedback system for data mining, the automatic feedback system for data mining includes a data segmentation module, a result evaluation module, and a parameter adjustment module; the data segmentation module is used to divide data into training data and evaluation data ; The result evaluation module is used to evaluate the satisfaction of the data mining results, and the evaluation results are fed back to the parameter data adjustment module; the parameter adjustment module adjusts the parameters of the data mining algorithm according to the evaluation of the result evaluation module.

A method for establishing an automatic feedback system for data mining involved in the present invention, the steps of which are as follows:

Step 1. Randomly divide the source data to be mined into training data and test data in proportion. The training data will be used to train the data mining algorithm model, and the test data will be used to evaluate the accuracy of the data mining model. Sub-segmentation and different random prototypes are used to avoid the impact of random segmentation on the evaluation of algorithm results;

Step 2. If the output of the data mining algorithm is a model, then use the independent variable of the test data generated by the data segmentation in step 1 as input, use the algorithm model generated by the data mining algorithm training for data mining, and compare the original data in the test data in step 1 Calculate the matching degree of the data results and the output of mining using the algorithm model, and calculate the network performance indicators such as MSE and RMSE on the matching to obtain the accuracy evaluation of the algorithm model;

If the output of the data mining algorithm is the result data, compare the data mining results generated by the training data with the test data, calculate the matching degree of the two, calculate the network performance indicators such as MSE and RMSE from the matching data, and compare the matching degree and The network performance index is fed back to the parameter data adjustment module;

Step 3. According to the test results of the data mining algorithm model and the accuracy evaluation of the algorithm model in step 2, and according to the feedback result of the result evaluation module, the parameters of the data mining are adjusted using the parameter automatic adjustment algorithm;

Step 4. Use the data mining algorithm model after parameter adjustment as a new algorithm model, and re-execute step 1 until the test results of the data mining algorithm model meet the requirements;

The parameter automatic adjustment algorithm described in step 3 includes: dividing the parameters into scalar parameters and vector parameters; when performing parameter adjustment, first adjust the scalar parameters, and when the adjustment of the scalar parameters still cannot meet the requirements, gradually change the granularity from coarse to fine Adjust each vector parameter;

Further, the above-mentioned scalar parameter refers to a parameter whose value is a limited number of values, such as the similarity distance method can only take limited values such as Euclidean distance, Minkowski distance, and Manhattan distance;

Further, the above-mentioned vector parameters refer to parameters that can be adjusted with any floating point number within a certain range, such as the smoothing parameters of the naive Bayesian classification algorithm;

Wherein the performance function of a kind of MSE network described in step 2 is the mean square error of the network, and its calculation method is as follows:

$\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; S</span>}\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r</span>}}$

Wherein the RMSE described in step 2 is a performance function of a network, which is the root mean square error of the network, and its calculation method is as follows:

$\mathrm{<span\; class="notranslate">\; R</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; S</span>}\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span>\sqrt{\frac{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; b</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; mod</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; no</span>}}}$

A data mining automatic feedback system established by the method of the present invention, its data segmentation module, result evaluation module, and parameter adjustment module work together to form feedback, automatically adjust and optimize the parameters of the data mining algorithm, and save more manpower than previous data mining cost. The data is divided into training data and test data through the data segmentation module, so that the data mining effect verification is evidence-based. The result evaluation module evaluates the results of the data mining algorithm and gives feedback on the data mining effect to make the parameter adjustment more scientific. The parameters of the data mining algorithm are automatically adjusted through the parameter adjustment module to reduce the waste of manpower caused by the use of expert experience models. Through the parameter automatic adjustment algorithm, the algorithm parameters can be adjusted efficiently and accurately, which can effectively improve the efficiency of parameter adjustment and reduce the actual waste caused by excessive traversal of parameter value ranges in the automation process.

Description of drawings

Fig. 1 is the working flowchart of the data mining automatic feedback system in the embodiment of the present invention;

Among them, 1 is the data segmentation module, 2 is the result evaluation module, and 3 is the parameter adjustment module.

Fig. 2 is the working flow chart of data segmentation module in the embodiment of the present invention;

Fig. 3 is the working flowchart of the result evaluation module in the embodiment of the present invention;

Fig. 4 is the working flowchart of the result evaluation module in the embodiment of the present invention;

Fig. 5 is a working flow diagram of the parameter adjustment module in the embodiment of the present invention.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Embodiment one

The embodiments described by referring to the figures are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

A data mining automatic feedback system involved in this embodiment, its work flow chart is shown in Figure 1, and the specific steps are as follows:

Step 1. Data segmentation

As shown in Figure 2, the source data to be mined is randomly divided into training data and test data in proportion. The training data will be used to train the data mining algorithm model, and the test data will be used to evaluate the accuracy of the data mining model. Perform multiple splits and use different random prototypes to avoid the impact of random splits on the evaluation of algorithm results.

Step 2. Train and evaluate data mining algorithms

As shown in Figure 3, if the output of the data mining algorithm is a model, the independent variable of the test data generated by the data segmentation in step 1 is used as input, and the algorithm model generated by the data mining algorithm training is used for data mining, and the test data in step 1 is compared Calculate the matching degree of the original data results and the output of mining using the algorithm model, and calculate the network performance indicators such as MSE and RMSE on the matching to evaluate the accuracy of the algorithm model.

As shown in Figure 4, if the data mining algorithm outputs result data, compare the data mining results generated by the training data with the test data, calculate the matching degree of the two, and calculate the network performance indicators such as MSE and RMSE from the matched data. And the matching degree and network performance index will be fed back to the parameter data adjustment module.

Wherein the performance function of a kind of MSE network described in step 2 is the mean square error of the network, and its calculation method is as follows:

$\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; S</span>}\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; no</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r</span>}}$

Wherein the RMSE described in step 2 is a performance function of a network, which is the root mean square error of the network, and its calculation method is as follows: $RmS\mathrm{E.}=\sqrt{\frac{{\mathrm{\&Sigma;}}_{i=1}^{\mathrm{no}}{(x_{ob\mathrm{the\; s},i}-x_{\mathrm{mod}el,i})}^2}{\mathrm{no}}}$

Step 3. Adjust the data mining algorithm parameters to optimize the data mining algorithm

As shown in Figure 5, according to the feedback result of the result evaluation module, the parameter automatic adjustment algorithm is used to adjust the parameters of data mining. The parameter automatic adjustment algorithm includes: divide the parameters into scalar parameters and vector parameters; when adjusting parameters, adjust the scalar parameters first. . Among them, the scalar parameter refers to the value of the parameter with a limited number of values. For example, the similarity distance method can only take limited values such as Euclidean distance, Minkowski distance, and Manhattan distance; the vector parameter refers to the value that can be used in a certain A parameter whose range can be adjusted in arbitrary floating point numbers, such as the smoothing parameter of the Naive Bayesian classification algorithm.

As shown in Figure 1, the data mining automatic feedback system includes three modules: data segmentation module 1, result evaluation module 2, and parameter adjustment module 3.

Among them, the result evaluation module 2, the parameter adjustment module 3 and the data mining algorithm form a feedback loop, and positive feedback optimization is continuously carried out in the feedback loop before satisfactory data mining results are obtained. The calculation result of the data mining algorithm or the obtained model is input to the algorithm evaluation module, the output result of the algorithm evaluation module is input to the parameter adjustment module, and the parameter adjustment result is applied to the data mining algorithm to form a circular operation system.

In the description of this specification, descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or feature is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications, substitutions and modifications can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the invention is defined by the claims and their equivalents.

Claims (1)

1. the method setting up data mining automatic feedback system, is characterized in that the method step is as follows:

Step 1, being training data and test data by source data to be excavated random division in proportion, wherein training data will be used for Training data mining algorithm model, test data for the accuracy of evaluating data mining model, for process each time perform into Row repeated segmentation and use different random prototype, it is to avoid because the occasionality of random division affects the evaluation to arithmetic result；

If step 2 data mining algorithm is output as model, then certainly becoming of test data data segmentation in step 1 produced Amount carries out data mining as input, the algorithm model using data mining algorithm training to produce, and tests data in comparison step 1 The data result of middle script and use algorithm model carry out the output excavated, and calculate the two matching degree, match and calculate MSE And the network performance index such as RMSE draws the Accuracy evaluation to algorithm model；

If data mining algorithm is output as result data, then data mining results training data produced and test data phase Relatively, calculating the two matching degree, the data matched calculate the network performance indexes such as MSE and RMSE, and just mate Degree and network performance index feed back to supplemental characteristic adjusting module；

Step 3, according in step 2 to data mining algorithm model test results and the Accuracy evaluation to algorithm model, according to The feedback result of evaluation of result module, uses the automatic adjustment algorithm of parameter to be adjusted the parameter of data mining；

Step 4, using adjusting the data mining algorithm model after parameter as new algorithm model, re-execute step 1, until number Requirement is reached according to the test result of mining algorithm model.