CN103235892A - Method for quickly quantizing parameter value distribution characters during manufacturing of plasma display panel - Google Patents

Abstract

The invention discloses a method for quickly quantizing parameter value distribution characters during manufacturing of a plasma display panel. The method comprises the steps of judging an input data source and statistical indicators through a computer, and dividing the data source according to working procedures and time; determining a maximum value, a minimum value, a mean value, a variance, skewness and kurtosis according to working procedures and the time, and producing a data distribution diagram according to the maximum value, the minimum value, the mean value, the variance, the skewness and the kurtosis; conducting classifying labeling on two data sets required to be compared, and then merging data of two classifications to obtain a data set with two classification labels; and then conducting important feature extraction and the like on the data set. The method has the advantages of being capable of quickly conducting quantitative comparison on observed values of parameters on the basis of a lot of production data, fundamentally solving the problems of difficult understanding, slow commencement and a long processing period on the condition of the complex and massive data and conducting effective information feedback timely for production states so as to facilitate quick searching and solving of problems for producers.

Description

The fast quantization method of parameter value distribution character in the plasma panel manufacture process

Technical field

The present invention relates to a kind of data mining technology of plasma panel manufacture process characteristic, especially relate to the fast quantization method of parameter value distribution character in the plasma panel manufacture process.

Background technology

So data complexity and data volume of article on plasma display screen face to face, apparatus volume data analytical approach (for example feature extraction, decision tree classification analysis, frequent rule digging etc.) often needs to expend a few minutes at least, and several hours time is carried out analytical calculation at most.When needs carried out a large amount of replica test, such time loss can become and be difficult to bear.

Summary of the invention

In order to overcome the above-mentioned shortcoming of prior art, the invention provides the fast quantization method of parameter value distribution character in a kind of plasma panel manufacture process, can represent a series of statistical indicator to the user, allow the user that the understanding of quantification can be arranged the distribution character of data fast.

The technical solution adopted for the present invention to solve the technical problems is: the fast quantization method of parameter value distribution character in a kind of plasma panel manufacture process comprises the steps:

Step 1, input data source and statistical indicator;

Step 2, computing machine are judged data source and the statistical indicator of input:

(1) when data source is the individual data collection, then enters step 3;

(2) when data source is two data sets, then enter step 5;

Step 3, according to operation and time data source is divided; Obtain maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then return step 2; If then enter step 4;

Step 4, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort shows and preserves; Return step 2 then;

Step 5, two data set dos classification that needs are compared mark, and merge the data of two classifications then, obtain a data acquisition with two types of marks; Carrying out key character at data acquisition again extracts;

Step 6, according to operation and time data source is divided; Calculate maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time in the important parameter set, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then data profile is shown in the list and preserves in contrast, return step 2 then; If then enter step 7;

Step 7, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort show in the contrast list and preserve, return step 2 then.

Compared with prior art, good effect of the present invention is: the present invention can compare in the quantification of a large amount of enterprising line parameter observed readings of production data fast, fundamentally solved in the face of complicated and mass data the time, it is difficult to understand, go into slow with one's hands and problem long processing period, can carry out the effective information feedback to production status in time, thereby provide shortcut for producers search fast and deal with problems.

Embodiment

The fast quantization method of parameter value distribution character comprises the steps: in a kind of plasma panel manufacture process

Step 1, input data source and statistical indicator:

Step 2, computing machine are judged data source and the statistical indicator of input:

(1) when data source is the individual data collection, then enters step 3;

(2) when data source is two data sets, then enter step 5;

Step 3, data source is divided according to operation and time (comprise by the sky and monthly); Obtain maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then return step 2; If then enter step 4;

Step 4, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort shows and preserves; Return step 2 then;

Step 5, two data set dos classification that needs are compared mark, and are about to a data set and are considered as (being treated to) classification, and another data acquisition is considered as (being treated to) another classification; The data that merge these two classifications then obtain a data acquisition with two types of marks; Carry out key character at this data acquisition then and extract, concrete key character abstracting method is as follows:

(1) input data set closed carry out that missing values is handled and normalized:

Described missing values is handled and is referred to: for missing values, fill according to the mean value of the observed reading of the identical parameters of other PDP display, target be the missing values that makes this filling to after analytical framework do not exert an influence;

After missing values is handled, observed reading on all parameters is carried out normalized, be about to the observed reading unification of this parameter on different PDP display and in this interval of 0 average and unit variance, guarantee simultaneously that also the relative size of this observed reading on the initial different screens is constant, illustrate as follows:

If α _i=(α _I1, α _I2..., α _In) representing a group observations of i parameter of n PDP display, the adjustment mode of new argument observed reading is after the normalization:

In the formula, μ _iBe α _iAverage, σ _iBe α _iVariance.

(2) adopt feature extracting method that the product of different grouping is carried out feature extraction, filter out the important parameter tabulation that can maximize the important evaluation index that comprises information gain, the ratio of gains, minimizing redundant degree and the maximization degree of correlation, concrete grammar is as follows:

Definition of T is a production data set, comprise (x, y)=(x ₁, x ₂..., x _n, y), x wherein _iThe value of i parameter in the PDP display data is produced in ∈ vals (t) representative, grade or the bad code of current this PDP display of y representative.

1) computing information gain (Information Gain):

$\mathrm{IG}(T,a)=H\left(T\right)-{\mathrm{\Σ}}_{v\∈\mathrm{vals}\left(t\right)}\frac{\left|\right\{x\∈T|x_i=v\}|}{\left|T\right|}\·H\left(\right\{x\∈T|x_i=v\}),$ Wherein

Be information entropy, p (x _i) representative observes x _iProbability, can be by occurring x in the historical data _iNumber of times calculate.

2) calculated gains is than (Gain Ratio):

IGR (T, a)=IG (T, a)/IV (T, a) wherein, IG (T a) is information gain, IV (T, a) account form is as follows:

$\mathrm{IV}(T,a)=-{\mathrm{\Σ}}_{v\∈\mathrm{vals}\left(t\right)}\frac{\left|\right\{x\∈T|x_i=v\}|}{\left|T\right|}\·{\log }_2\left(\frac{\left|\right\{x\∈T|x_i=v\}|}{\left|T\right|}\right);$

3) calculate maximal correlation degree and minimum redundancy (minimum Redundancy and Maximum Relevancy, mRMR)

A) computational minimization redundance (mR):

$\min W_I{,W}_I=\frac{1}{{\left|T\right|}^2}\underset{a,b\∈T}{\mathrm{\Σ}}I(a,b)$

Wherein, $I(a,b)={\mathrm{\Σ}}_{i,j}p(a_i,b_j)\·\log \frac{p(a_i,b_j)}{p\left(a_j\right)p\left(b_j\right)},$ Be defined as the interactive information between two variable a and the b;

B) calculate the maximization degree of correlation (MR):

$\max V_I,V_I=\frac{1}{\left|T\right|}\underset{a\∈T}{\mathrm{\Σ}}I(y,a)$

C) calculating the mode can satisfy mR and MR to the full extent is that method for searching by incremental finds the feature (parameter) that satisfies following condition:

maxΦ(W _I,V _I),Φ=V _I-W _I

Suppose that existing key character (parameter) set is S _M-1, wherein having comprised m-1 key character (parameter), target is in order to find m key character (parameter) to make it can maximize Φ from residue character (parameter) set.

More than the computing method of each feature selecting can export one group of key character (parameter), namely by IG, three stack features (parameter) that IGR, mRMR export respectively characterize feature (parameter) combination of discrimination is arranged in the current PDP display grouping most.

(3) the three groups of important parameters tabulation that (2) is obtained is integrated, and obtains out one group of unified important parameter tabulation, comprising appear at parameter at least two groups in three groups of key characters tabulations.

Step 6, data source is divided according to operation and time (comprise by the sky and monthly); Calculate maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time in the important parameter set, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then data profile is shown in the list and preserves in contrast, return step 2 then; If then enter step 7;

Step 7, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort show in the contrast list and preserve, return step 2 then.

Claims (4)

1. the fast quantization method of parameter value distribution character in the plasma panel manufacture process is characterized in that: comprise the steps:

Step 1, input data source and statistical indicator;

Step 2, computing machine are judged data source and the statistical indicator of input:

(1) when data source is the individual data collection, then enters step 3;

(2) when data source is two data sets, then enter step 5;

Step 3, according to operation and time data source is divided; Obtain maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then return step 2; If then enter step 4;

Step 4, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort shows and preserves; Return step 2 then;

Step 5, two data set dos classification that needs are compared mark, and merge the data of two classifications then, obtain a data acquisition with two types of marks; Carrying out key character at data acquisition again extracts;

Step 6, according to operation and time data source is divided; Calculate maximal value, minimum value, average, variance, the degree of bias, kurtosis by operation and time in the important parameter set, and make data profile accordingly; Continue then to judge whether to have statistical indicator: if not, then data profile is shown in the list and preserves in contrast, return step 2 then; If then enter step 7;

Step 7, each result is sorted by statistical indicator, and preceding K attribute of the gained that will sort show in the contrast list and preserve, return step 2 then.

2. the fast quantization method of parameter value distribution character in the plasma panel manufacture process according to claim 1, it is characterized in that: described key character abstracting method is:

(1) input data set is closed carries out that missing values is handled and normalized;

(2) filter out the parameter list that can maximize three groups of key characters that comprise information gain, the ratio of gains, minimizing redundant degree and the maximization degree of correlation;

(3) parameter list of three groups of key characters is integrated, obtained one group of unified important parameter tabulation, comprising in three groups of key characters tabulations, appearing at parameter at least two group key characters.

3. the fast quantization method of parameter value distribution character in the plasma panel manufacture process according to claim 2, it is characterized in that: described missing values is handled and is referred to: for missing values, fill according to the mean value of the observed reading of the identical parameters of other PDP display.

4. the fast quantization method of parameter value distribution character in the plasma panel manufacture process according to claim 2, it is characterized in that: described normalized refers to: with the observed reading unification of each parameter on different PDP display to 0 average and unit variance this interval in and guarantee that simultaneously the relative size of this observed reading on initial different the screen is constant.