CN103164499A

CN103164499A - Order clustering method during product planning

Info

Publication number: CN103164499A
Application number: CN2012101063390A
Authority: CN
Inventors: 孙永国; 张晓明; 葛江华
Original assignee: Harbin University of Science and Technology
Current assignee: Harbin University of Science and Technology
Priority date: 2012-04-12
Filing date: 2012-04-12
Publication date: 2013-06-19

Abstract

The invention provides an order clustering method during product planning. The method comprises the steps of standardizing sample data, establishing a fuzzy similar matrix, calculating fuzzy clustering transitive closure of the fuzzy similar matrix, choosing confidence level values suitably to finish clustering, conducting membership calculation to a series of standard values or standard intervals, determined by manufacturers and corresponding to all attribute values of each class of orders, of an attribute, regarding the largest value or the interval of the membership as the attribute value in the class of the orders, and conducting satisfaction calculation of clients in the class of the orders. The order clustering method during the product planning can greatly reduce time of product clustering, improve efficiency of the product planning, process the order clustering accurately and fast, improve satisfaction of the clients, and lower production cost.

Description

Order clustering method in a kind of product planning

Technical field

The present invention relates to a kind of order clustering method, espespecially a kind ofly shorten the whole product planning time not affecting under the prerequisite of integral product planning, meet customer need to greatest extent to carry out the fuzzy clustering method of next step product coupling work.

Background technology

Cluster analysis is one group of set of records ends of not demarcating of input, and namely the record of input also is not carried out any classification, then according to certain rule and methodology, classifying rationally set of records ends.In other words, so-called cluster is exactly that certain attribute according to things is divided into some composition classifications with a group objects, makes similarity be in same group greater than the object elements of setting value, is distributed in the process of different groups less than the object elements of this value.In like manner, the cluster of customer demand refers to take the customer demand that gets by variety of way as object, calculates by the similarity between the attribute item of demand demand is divided, thereby draw different classes of requirements set.

In existing technology, the production run of product enters the product configuration from the input product order, namely mates with existing product, then enters the production scheduling process of product, exports at last finished product, three processes of front and back experience.For example, when 500 products are produced, should carry out the product configuration of 500 times, the production scheduling process of 500 times according to step.

Can find out from said process, existing product planning does not have this process of cluster, has so just virtually increased the production cycle of product, has reduced production efficient, enterprise also can reduce on benefit to some extent simultaneously, is difficult in time meet customer need.

In a word, need at present the technical matters that those skilled in the art solves to be exactly: a kind of product planning method of Innovation, make whole process of producing product, can satisfy enterprise's demand on benefit, also can satisfy the client and obtain at short notice oneself satisfied product.

Summary of the invention

Technical matters to be solved by this invention is, a kind of clustering method is provided, and namely adds in the product planning process, makes with short production cyclely in process of producing product, and enterprise's productivity effect is high, satisfies fast client's individual demand.

Clustering method of the present invention comprises: the normalization sample data; Set up fuzzy similarity matrix; Calculate the fuzzy clustering transitive closure of fuzzy similarity matrix; Suitably choose confidence value and complete cluster; Calculate carrying out degree of membership between the series of standards value of this attribute of being determined by manufacturer corresponding to each property value in every class order or standard regions again after cluster, and with the value of degree of membership maximum or interval as this property value in such order, carry out at last the satisfaction of client in the class order and calculate.

Its iterative process is when the computing client satisfaction: selected satisfied

, establish Initial value be

, and satisfy

, iteration step length is , when

, finishing iteration,

Be a very little positive number, its iterative process is as follows:

Step1: order

, , carry out iteration

, with

For threshold values carries out the demand cluster, and the computing client satisfaction

Step2: differentiate customer satisfaction, if

Turn step 1, if

Turn Step 3;

Step3: order ,

, with

Step4: differentiate customer satisfaction, if , and

Turn step 5; If

, and

, turn Step 3; If Turn Step 6;

Step5: Get this iterative value, finish;

Step6:

Get last iterative value, finish.

Carry out iteration by above step and can realize conflict optimization between customer satisfaction and cluster granularity making cluster result namely of certain scale, can make again the client have suitable satisfaction.

In clustering algorithm in step 1, be fuzzy through the order specific requirement after cluster, in order the design personnel to be configured design, with fuzzy demand accuracy.

In clustering algorithm in step 1,

Fuzzy similarity matrix just, only having when R is fuzzy equivalent matrix could cluster, therefore R need to be transformed into fuzzy equivalent matrix.Transmit bag n rank fuzzy similarity matrix R is transformed into n rank fuzzy equivalent matrix by asking

In clustering algorithm in step 1, after fuzzy customer demand precision, may produce deviation with client's demand intention, need to determine that this deviation whether within client's acceptable scope, need to verify customer satisfaction this moment.If the client is dissatisfied as a result, illustrate that the cluster granularity is too large, need to increase threshold values, to reduce the cluster granularity, increase customer satisfaction.

In this application, adopt the method for fuzzy clustering to carry out cluster to customer order, can further reduce the burden of system resource, the use amount of CPU and internal memory for example, has improved the operational performance of clustering method at the working time of having reduced cluster.

The invention provides a kind of method of order cluster, add cluster process in the product planning process, compare product planning process in the past, with short production cycle, enterprise's productivity effect is high, and product is saved time in the finished product process that places an order, raise the efficiency, satisfy fast client's individual demand.

Description of drawings

Fig. 1 is the clustering method schematic diagram;

Fig. 2 is customer satisfaction algorithm calculation flow chart;

Fig. 3 is that each demand similarity is calculated classification chart;

Fig. 4 represents the form interface for forming customer order;

Fig. 5 is the requirement specification interface;

Fig. 6 is for calculating the similarity interface of any two orders;

Fig. 7 is input parameter coefficient of similarity interface;

Fig. 8 is for setting up the similar matrix interface;

Fig. 9 is input threshold values interface;

Figure 10 is that threshold values calculates and cuts the matrix interface;

Figure 11 is preservation cluster result interface.

Embodiment

Below in conjunction with accompanying drawing with use the product planning system of this clustering method to make further more detailed description to technical scheme of the present invention.

1, implement the basis and prepare, the order input

Suppose the order that existing four clients propose, as shown in the table:

2, the order expression way is defined

As accompanying drawing 3, the order data after input is arranged by requirement item.

3, requirements normalization

As accompanying drawing 4, the expression way of definition is standardized.Adopt clustering method that every demand is quantized, form matrix form.

4, calculate the similarity of any two orders

As accompanying drawing 5, the normalizated value by previous step is found the solution calculates the similarity that any two kinds of orders carry out under demand relation.

5, input parameter coefficient of similarity

As accompanying drawing 6, namely input weight coefficient, the importance degree of a requirement item in other requirement items can be changed by leading subscriber herein voluntarily.

6, set up similar matrix

As accompanying drawing 7, certain order and set up the similar matrix of a symmetry between relevant, " 1 " at diagonal angle refers to that certain order or related and itself similarity are 1.

7, input threshold values

As accompanying drawing 8, input threshold values on the basis that forms similar matrix, the Different Effects of threshold values is to the result of order with related coupling.

8, calculate a section matrix by threshold values

As accompanying drawing 9, the not success of " 0 " expression coupling, the match is successful in " 1 " expression.

9, saving result

As accompanying drawing 10.

Claims

1. the order clustering method in a product planning, its composition comprises: the normalization sample data; Set up fuzzy similarity matrix; Calculate the fuzzy clustering transitive closure of fuzzy similarity matrix; Suitably choose confidence value and complete cluster; Calculate carrying out degree of membership between the series of standards value of this attribute of being determined by manufacturer corresponding to each property value in every class order or standard regions again after cluster, and with the value of degree of membership maximum or interval as this property value in such order, carry out at last the satisfaction of client in the class order and calculate.

2. clustering method as claimed in claim 1, is characterized in that, adopts the normalized method of logarithm in the normalization sample data.

3. clustering method as claimed in claim 1, is characterized in that, the similarity calculating in similar matrix between any two demands is the similarity by the demand type , the demand name similarity

, the requirements type similarity Similarity with requirements Four parts form.

4. clustering method as claimed in claim 1, is characterized in that, adopts the result of the method judgement order cluster of the suitable threshold values of input.

5. clustering method as claimed in claim 1, is characterized in that, comes the computing client satisfaction with process of iteration.

6. clustering method as claimed in claim 1, is characterized in that, the order specific requirement that the client is proposed is configured, with fuzzy demand accuracy.

7. clustering method as claimed in claim 1, is characterized in that, wraps by transmission traditional similar matrix is transformed into fuzzy equivalent matrix, uses fuzzy equivalent matrix to carry out cluster.

8. clustering method as claimed in claim 1, is characterized in that, precision is reasonably revised and made it to the deviation that produces after cluster, by the minor modifications of threshold values being improved client's satisfaction, finally reaches customer satisfaction and maximize.

?