CN101430735B - Protective farming mode selection method - Google Patents

Abstract

The invention relates to a method for selecting a protective farming mode, which is characterized by comprising the following steps: 1) data collection: soil data, climate data, geographic data and production habit data of a mature area and an area to be implemented in different modes are collected; 2) preparation of mode selection: a decision unit and an index of each protective farming mode are determined according to the data of the mature area; 3) construction of a projection pursuit analysis model; 4) construction of a similarity evaluation model; 5) model selection: the protective farming mode corresponding to the mature area with the highest similarity to the area to be implemented is selected as the protective farming mode of the area to be implemented according to the level of similarity. The invention selects a proper protective farming mode for the area to be implemented by constructing a protective farming mode selection mode, and the whole process is completed by computer analog simulation without field experiment, thereby improving the efficiency of protective farming mode selection, and reducing the cost for popularizing protective farming.

Description

A kind of protective farming mode selection method

Technical field

The present invention relates to a kind of mode identification technology, particularly about a kind of protective farming mode selection method that is applied in the agricultural production.

Background technology

Conservation tillage is a kind of novel dry farming, and it mainly comprises no-tillage seeding fertilization, dark pine, four contents such as control weeds, stalk and soil surface treatment.Its core is a no-tillage seeding, and key operation all uses machinery to finish.Protective farming technique originates from the U.S. of the thirties in 20th century, and states such as the U.S., Canada and Australia have substantially all adopted the conservation tillage that turns to support with machinery at present, have obtained effect preferably.

Conservation tillage has only the more than ten years in the time that China carries out system test and research, has obtained certain achievement in some type areies, has summed up the pattern of a series of conservation tillages of suitable particular locality.Have nothing in common with each other because regional soil, weather, geography, crop species, production custom etc. are implemented in different waiting, therefore can not adopt identical conservation tillage pattern them.So, in the process of promoting protective farming technique, in the past mainly by carry out the pattern that the field experiment screening is fit in area to be implemented; though the result that this method obtains is more directly perceived; but the length of expending time in, the input cost height has limited the kind of pattern and the scale of test.

Summary of the invention

At the problems referred to above; the purpose of this invention is to provide a kind of under non-experiment condition; by making up conservation tillage model selection model; compare the ripe area of conservation tillage and wait to implement interlocal similarity, for waiting to implement the regional quick suitable protective farming mode selection method of reasonably selecting.

For achieving the above object, the present invention takes following technical scheme: a kind of protective farming mode selection method is characterized in that: it may further comprise the steps: 1) data aggregation: soil data, climatic data, geodata, the production custom data of collecting the ripe area of different mode; Collection waits to implement soil data, climatic data, geodata, the production custom data in area; 2) model selection is prepared: according to the data in described ripe area, determine the decision package and the index of each conservation tillage pattern; 3) make up the projection pursuit analytical model: detection, the enquiry data in ripe area are carried out standardization,, determine the objective weight of each index with projection pursuit technique according to described decision package and index ${\mathrm{\β}}_j^{\left(1\right)}=\{{\mathrm{\β}}_1^{\left(1\right)},{\mathrm{\β}}_2^{\left(1\right)}...{\mathrm{\β}}_p^{\left(1\right)}\},$ Wherein p is the number of index; 4) make up the similarity evaluation model: calculate the index similarity complex weight β that waits to implement between regional detection, enquiry data and each different regions _jCalculate the similarity value of waiting to implement between area and the ripe area afterwards, estimate the similarity and the similarity grade in these two areas; 5) model selection:, select to wait to implement the regional pairing conservation tillage pattern of the highest described maturation of regional similarity, as the conservation tillage pattern of waiting to implement the area with described according to described similarity grade.

In the described step 1), the soil in described ripe area is data from existing experimental result and on-site inspection result; The described soil The data chemical gauging of waiting to implement the area; Described maturation is regional and wait that the climatic data of implementing the area obtains according to 10 years meteorological data weighted means; Described ripe area obtains with the The data field method of the geodata of waiting to implement the area, production custom.

Described step 2) in, the classification in the corresponding ripe area of described decision package, described index derive from soil data, climatic data, geodata, the production custom index that has comparability in the ripe area.

In the described step 3), make up described projection pursuit analytical model and may further comprise the steps:

A) data normalization: with raw data matrix X ^*Be converted into standardization matrix X, x _IjBe the element among the standardization matrix X, represent j desired value of i decision package;

B) linear projection: establish vector of unit length a={a ₁, a ₂... a _pBe the one-dimensional linear projecting direction, then to project to the one dimension projection properties value on the projecting direction a be z to standardization matrix X _i

$z_i=\underset{j=1}{\overset{p}{\mathrm{\Σ}}}{a}_j\·x_{\mathrm{ij}}$

C) structure projection target function:

Q(a)＝S _Z·D _Z

$S_z=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(z_i-E\left(z\right))}^2}{n-1}}$

$D_z=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}(R-r_{\mathrm{ik}}))\·f(R-r_{\mathrm{ik}})$

Wherein, E (z) is sequence { z _i| i=1 ..., the mean value of n|}; If r _IkBe the distance between the projection properties value; R is the density window width, and its span is: $\max \left(r_{\mathrm{ij}}\right)+\frac{p}{2}\≤R\≤2p;$ r _ik＝|z _i-z _k|，i，k＝1，…，n；

$f\left(t\right)=\left\{\begin{array}{cc}0& t\&GreaterEqual;0\\ 1& t<0\end{array}\right.;$

D) optimize projection target function Q (a): the maximization of utilization objective function is optimized projecting direction a the projection target function, make projection target function Q (a) reach maximum value projecting direction be the best projection direction

:

MaxQ(a)＝S _z·D _z

$\mathrm{sub}.\mathrm{to}:\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a_j}^2=1$

E) objective weight is calculated: the objective weight of determining described each index

${\mathrm{\&beta;}}_j^{\left(1\right)}=\frac{a_j^{*}}{\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a}_j^{*}}.$

In the described step 4), the step that makes up described similarity evaluation model is as follows: i) index classification; Ii) calculate index similarity S between different regions (A, B), A and B are two different regions; Iii) index similarity complex weight is definite:

${\mathrm{\&beta;}}_j=f({\mathrm{\&beta;}}_j^{\left(1\right)},{\mathrm{\&beta;}}_j^{\left(2\right)})=\frac{{\mathrm{\&beta;}}_j^{\left(1\right)}\&CenterDot;{\mathrm{\&beta;}}_j^{\left(2\right)}}{\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_j^{\left(1\right)}\&CenterDot;{\mathrm{\&beta;}}_j^{\left(2\right)}}$

Wherein, described ${\mathrm{\&beta;}}_j^{\left(1\right)}=\{{\mathrm{\&beta;}}_1^{\left(1\right)},{\mathrm{\&beta;}}_2^{\left(1\right)}...{\mathrm{\&beta;}}_p^{\left(1\right)}\}$ Try to achieve by described projection pursuit analytical model; ${\mathrm{\&beta;}}_j^{\left(2\right)}=\{{\mathrm{\&beta;}}_1^{\left(2\right)},{\mathrm{\&beta;}}_2^{\left(2\right)},...{\mathrm{\&beta;}}_p^{\left(2\right)}\},$ The parameter that the behaviour wage adjustment is whole; Iv) interzone similarity evaluation: regional A comprises φ index, and regional B comprises

Individual index, the coincident indicator of regional A and B have σ,

The similarity S between j the similar index wherein _iBy ii) the step draws; β _jThe complex weight coefficient of the j that the iii) step of serving as reasons is obtained a similar unit; Similarity S between area A and the B ^*(A, mathematical model B):

$S_u={\mathrm{\&beta;}}_1\&CenterDot;S_1+{\mathrm{\&beta;}}_2\&CenterDot;S_2+...{\mathrm{\&beta;}}_{\mathrm{\&sigma;}}\&CenterDot;S_{\mathrm{\&sigma;}}=\underset{j=1}{\overset{\mathrm{\&sigma;}}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_j\&CenterDot;S_j.$

P index is divided into according to the difference of estimating method in the described step 3): determined value index, determined value and interval value index, binary variable index, nominal data index and qualitative index.

The present invention is owing to take above technical scheme; it has the following advantages: 1, the present invention by with the ripe area of conservation tillage and successful pattern thereof as standard form; make up conservation tillage model selection model; for waiting that implementing the area selects suitable conservation tillage pattern; whole process is all finished by computer simulation; need not to carry out field experiment, thereby improved the efficient of conservation tillage model selection, reduced the cost that conservation tillage is promoted.2, the present invention is according to the experimental data in the ripe area of the conservation tillage of being gathered, determine decision package and index, structure is treated the execution area and is carried out similarity evaluation, thereby determine to adapt to the conservation tillage pattern of this area based on the conservation tillage model selection model of similarity; Conservation tillage model selection model of the present invention adopts distinct methods to determine each index similarity according to the difference of different index properties, has analytical approach advanced person, result advantage accurately.What protective farming mode selection method of the present invention can be different weathers, geography, production custom waits that implementing the area selects suitable conservation tillage pattern.

Description of drawings

Fig. 1 is a protective farming mode selection method synoptic diagram of the present invention

Fig. 2 is a structure projection pursuit analytical model process flow diagram of the present invention

Fig. 3 is a structure similarity evaluation model process flow diagram of the present invention

Embodiment

Below in conjunction with drawings and Examples the present invention is described in detail.

As shown in Figure 1, protective farming mode selection method provided by the invention may further comprise the steps:

1) data aggregation.Different at the facility of ripe area feature of conservation tillage and use, technology etc. are divided into different conservation tillage patterns with them, are gathered into soil data, climatic data, geodata, production custom and other significant data in cultivated land district.Conservation tillage area to be implemented is detected, investigated, collect these regional soil data, climatic data, geodata, production custom and other significant data.

2) model selection is prepared.According to detection, the enquiry data in ripe area, determine the decision package and the index of each conservation tillage pattern.

3) make up the projection pursuit analytical model.At first detection, the enquiry data in ripe area are carried out standardization; Secondly according to step 2) in decision package and the index determined, determine the objective weight of each index with projection pursuit technique ${\mathrm{\&beta;}}_j^{\left(1\right)}=\{{\mathrm{\&beta;}}_1^{\left(1\right)},{\mathrm{\&beta;}}_2^{\left(1\right)}...{\mathrm{\&beta;}}_p^{\left(1\right)}\},$ Wherein p is the number of index.

4) make up the similarity evaluation model.Calculating waits to implement the index similarity complex weight β between regional detection, enquiry data and each different regions _jCalculate the similarity value of waiting to implement between area and the ripe area afterwards, estimate the similarity and the similarity grade in these two areas.

5) model selection.According to the similarity grade, select and wait to implement the high ripe regional pairing conservation tillage pattern of regional similarity, as the conservation tillage pattern of waiting to implement the area.

In the inventive method step 1), at first different according to the facility of ripe area feature of conservation tillage and use, technology etc. are distinguished them.Such as: present embodiment zoning maturely is divided into seven different areas, each area corresponding one or more conservation tillage pattern (only as example, but being not limited thereto):

1, no-tillage whole stalk covering, the whole stalk of shallow pine cover;

2, no-tillage broken stalk covering, the shallow stalk that ravels cover;

3, no-tillagely stay that low stubble covers, shallow pine stays low stubble to cover;

4, no-tillage high stubble covering, the shallow Song Liugao stubble of staying covers;

5, the stalk that ravels deeply covers;

6, no-tillage fixedly road;

7, no-tillage ridge culture.

The present invention calculates the similarity of waiting to implement between area and above seven the ripe areas respectively, selects rational conservation tillage pattern.Before modeling, collect and be organized into the cultivated land district and wait to implement the experimental data in area, comprise soil data, climatic data, geodata, production custom and other significant data.The soil in ripe area is data from existing experimental result and on-site inspection result, wait that the soil data of implementing the area are recorded by following chemical method: core cutter method measured soil unit weight, potassium bichromate titrimetric method measured soil organic matter, constant temperature oven oven drying method measured soil water cut, the full nitrogen of Kelvin meso method measured soil, diffusion method measured soil alkali-hydrolyzable nitrogen, molybdenum antimony scandium colourimetry measured soil available phosphorus, flare photometer measured soil available potassium is with the pH value of PH instrumentation amount soil.Ripe area obtains according to 10 years meteorological data weighted means with climatic datas such as treating the average annual quantity of precipitation in execution area, average annual temperature.The field method acquisition is adopted with the geodata of waiting to implement the area, production custom and other significant data in ripe area.

The inventive method step 2) in; the decision package of conservation tillage pattern and index are according to the enquiry data in ripe area is carried out data mining; determine by the data messages such as relation schema between research measuring method, the field of investigation and the survey region; the classification in the corresponding ripe area of decision package wherein, index derives from soil data, climatic data, geodata, production custom and other significant data that has comparability in the ripe area.In the present embodiment, decision package number n=7, each decision package comprises 1 ripe area, therefore always has 7 ripe areas.Index number p=20 is (only as example, but be not limited thereto), 20 indexs are respectively: the soil weight, water cut, organic matter, full nitrogen, alkali-hydrolyzable nitrogen, available phosphorus, available potassium, pH value, the soil texture, average annual quantity of precipitation, average annual temperature, accumulated temperature, sunshine time, the gradient, length of grade, longitude, dimension, shortening type, crop species and irrigation conditions.

As shown in Figure 2, in the inventive method step 3), the step that makes up the projection pursuit analytical model is as follows:

A) data normalization.At first determine that according to detection, the enquiry data in ripe area 7 decision packages and 20 indexs constitute raw data matrix X ^*, with raw data matrix X ^*Be converted into standardization matrix X, with the difference of eliminating each index dimension and the influence of magnitude difference.X wherein _IjBe the element among the standardization matrix X, j desired value of i decision package after the expression standardization.Embodiments of the invention are distinguished their standardization strategy according to the practical significance of index.In agricultural production, high more being easy to more of crop yield accepted by the peasant, and then this class index of crop yield belongs to benefit type index; Few more being easy to more of fuel consume accepted by the individual dealer of agricultural machines, and then this class index of fuel consume belongs to cost type index; The pH value of soil approaches intermediate value 7 more, and soil attribute is good more, and this class index of the pH value of soil belongs to the fixed index, also is called the index of intermediate value optimum.The standardized method that the index of different meanings is carried out is:

For the big more excellent more benefit type index of numerical value, adopt following processing mode:

$x_{\mathrm{ij}}=\frac{{x^{*}}_{\mathrm{ij}}-\min \left(x_j\right)}{\max \left(x_j\right)-\min \left(x_j\right)}$

For the more little excellent more cost type index of numerical value, adopt following processing mode:

$x_{\mathrm{ij}}=\frac{\max \left(x_j\right)-{x^{*}}_{\mathrm{ij}}}{\max \left(x_j\right)-\min \left(x_j\right)}$

For the index of intermediate value optimum, handle in the following ways:

$x_{\mathrm{ij}}=\frac{|{x^{*}}_{\mathrm{ij}}-\mathrm{mid}\left(x_j\right)|}{\mathrm{mid}\left(x_j\right)}$

In the formula: i is a decision package; J is an index.

Max (x _j) maximal value of j index of-----Di

Min (x _j) minimum value of j index of-----Di

Mid (x _j) intermediate value of j index of-----Di

When the number of samples of j index was odd number, peek value sample value placed in the middle was mid (x _j); When the number of samples of j index was even number, the mean value of two sample values that the peek value is placed in the middle was mid (x _j).

B) linear projection.Be about to original high dimensional data x _IjProjecting to the one-dimensional linear space handles.Because original index has p, and the p dimension data is spatially observed difficulty, can't reflect the true characteristics of data, and is easy to generation " dimension disaster ".As original high dimensional data x _IjAfter projecting to the one-dimensional linear space, the projection value that comes down to these p dimension indicator data that observes, projection value has reflected original index data characteristic, intuitive and convenient.If vector of unit length a={a ₁, a ₂... a _PBe the one-dimensional linear projecting direction, then to project to the one dimension projection properties value on the projecting direction a be z to standardization matrix X _i

$z_i=\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a}_j\&CenterDot;x_{\mathrm{ij}}$

C) structure projection target function.Make original high dimensional data x _IjBetween class distance S in one-dimensional space distribution _zWith density D in the class _zObtain maximal value simultaneously.Then the projection target function representation is:

Q(a)＝S _Z·D _Z

$S_z=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(z_i-E\left(z\right))}^2}{n-1}}$

$D_z=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}(R-r_{\mathrm{ik}}))\&CenterDot;f(R-r_{\mathrm{ik}})$

Wherein, E (z) is sequence { z _i| i=1 ..., the mean value of n|}, n are the number of decision package, p is the number of index.If r _IkBe the distance between the projection properties value, R is the density window width, and its span is: $\max \left(r_{\mathrm{ij}}\right)+\frac{p}{2}\&le;R\&le;2p.$ r _ik＝|z _i-z _k|，i＝ ₁，…，n；k＝ ₁，...，n。 $f\left(t\right)=\left\{\begin{array}{cc}0& t\&GreaterEqual;0\\ 1& t<0\end{array}\right..$

D) optimize projection target function Q (a).For given sample set desired value, projection target function Q (a) changes along with the variation of projecting direction a, can make projection target function Q (a) reach maximum value projecting direction be the best projection direction

Therefore use the objective function maximization that the projection target function is optimized projecting direction a:

MaxQ(a)＝S _z·D _z

$\mathrm{sub}.\mathrm{to}:\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a_j}^2=1$

E) objective weight

Calculating.Best projection direction according to step d) optimization

The value size is arranged index, utilizes following formula can determine the objective weight of each index It has reflected the contribution of the contained quantity of information of index to evaluation result itself.

${\mathrm{\&beta;}}_j^{\left(1\right)}=\frac{a_j^{*}}{\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{a}_j^{*}}$

As shown in Figure 3, in the step 4) of the inventive method, the step that makes up the similarity evaluation model is as follows:

I) index classification.Estimate the difference of method according to index, they are belonged to different classifications, so that according to the index similarity between all kinds of property calculation different regions.Present embodiment with the desired value in the ripe area of conservation tillage as standard value, being classified as follows of above-mentioned 20 indexs (as shown in table 1):

Table 1: the classification of index

Numbering	Classification	Index
			1	The determined value index	The soil organism, full nitrogen, alkali-hydrolyzable nitrogen, available phosphorus, available potassium, longitude, latitude
2	Determined value and interval value index	P in soil H, water cut, unit weight
			3	The interval value index	Average annual quantity of precipitation, average annual temperature, accumulated temperature, sunshine time, the gradient, length of grade
4	The binary variable index	The shortening type
			5	The nominal data index	Crop species
6	Qualitative index	The soil texture, irrigation conditions

II) the index similarity between the calculating different regions.According to index similarity characteristics in the table 1, adopt distinct methods to calculate the similarity of heterogeneity index.If two different regions A and B, their a certain common index L (A) and L (B), and the similarity S of two indexes (A, B), and S (A, B) ∈ [0,1].

I) determined value index similarity is calculated.When obtaining determined value behind the step a) data normalization in the index process step 3) of regional A and B, index L (A), L (B) ∈ [0,1].Similarity between the determined value can represent that distance is near more with the distance of point-to-point transmission, and similarity is high more.Then have:

$S(A,B)=1-|L\left(A\right)-L\left(B\right)|=\left\{\begin{array}{cc}1-(L\left(A\right)-L\left(B\right)),& L\left(A\right)\&GreaterEqual;L\left(B\right)\\ 1-(L\left(A\right)-L\left(B\right)),& L\left(A\right)<L\left(B\right)\end{array}\right.$

Ii) determined value and interval value index similarity are calculated.When the index L in two areas (A) and L (B) all be through obtain after the standardization interval the time, L (A)=[a ₁, a ₂], L (B)=[b ₁, b ₂], a ₁＜a ₂, b ₁＜b ₂, and a ₁, a ₂, b ₁, b ₂∈ [0,1].Similarity between determined value and the interval, relevant with the distance and the length of an interval degree of determined value and interval mid point respectively, the distance of determined value and interval mid point is more little, and the length of an interval degree is short more, and similarity is high more.

$S(A,B)=[1-(b_2-b_1)]\&CenterDot;[1-|L\left(A\right)-\frac{b_1+b_2}{2}|]$

$=\left\{\begin{array}{cc}[1-(b_2-b_1)]\&CenterDot;[1-(L\left(A\right)-\frac{b_1+b_2}{2})],& L\left(A\right)\&GreaterEqual;\frac{b_1+b_2}{2}\\ [1-(b_2-b_1)]\&CenterDot;[1-(\frac{b_1+b_2}{2}-L\left(A\right))],& L\left(A\right)<\frac{b_1+b_2}{2}\end{array}\right.$

Iii) interval value index similarity is calculated: the similarity between two intervals, and adopt interval level of coverage to represent, level of coverage is big more, and similarity degree is high more.

$S(A,B)=\mathrm{\&beta;}\&CenterDot;\frac{T_x\left(\mathrm{AB}\right)}{T\left(A\right)}+(1-\mathrm{\&beta;})\&CenterDot;\frac{T\left(A\right)}{T_{\max }\left(\mathrm{AB}\right)}$

$=\mathrm{\&beta;}\&CenterDot;\frac{T_x\left(\mathrm{AB}\right)}{T\left(B\right)}+(1-\mathrm{\&beta;})\&CenterDot;\frac{T\left(B\right)}{T_{\max }\left(\mathrm{AB}\right)}$

Wherein, T _x(AB) be the length of interval index L (A) and L (B) lap, and T (A)=| a ₂-a ₁|, T (B)=| b ₂-b ₁|; T _Max(AB)=max (a ₁, a ₂, b ₁, b ₂)-mim (a ₁, a ₂, b ₁, b ₂).β ∈ [0,1] is a weight, composes power with manual method.

Iv) binary variable index similarity is calculated: binary variable represents to have only 0 or 1 two state variable in this index.If it all is 1 variable number that h represents index L (A) and L (B), f represent index L (A) be 1 and index L (B) be 0 variable number, s represent index L (A) be 0 and index L (B) be 1 variable number, t represents that index L (A) and L (B) are 0 variable number, d (A, B) distinctiveness ratio of expression index L (A) and L (B), i.e. their Euclidean distance.Then the similarity of binary variable is:

$S(A,B)=1-d(A,B)=1-\frac{f+s}{h+f+s+t}.$

V) nominal data index similarity is calculated: nominal data is the popularization of binary variable, represents in the index that promptly the variable of state is more than two.Suppose that m represents the number that variable overlaps among index L (A) and the L (B), q represents the number of the whole variablees of two indexes, d (then the nominal data similarity is for A, the B) distinctiveness ratio of expression index L (A) and L (B):

$S(A,B)=1-d(A,B)=1-\frac{q-m}{q}=\frac{m}{q}.$

Vi) the qualitative index similarity is calculated: the similarity of qualitativing concept, usually will bluring qualitatively, notion is mapped to the numerical value description, comparison by numerical value utilizes step I) in the similarity formula of two determined value indexs, thereby realize the qualitative of similarity to Quantitative yield.For example, with irrigation conditions situation { good, better, general, relatively poor, poor } and numerical value 1,0.8,0.6,0.4, set up mapping relations between 0.2}.

III) index similarity complex weight β _jDetermine.Index similarity complex weight is the similarity that is used for the same index of definite two different regions.

${\mathrm{\&beta;}}_j=f({\mathrm{\&beta;}}_j^{\left(1\right)},{\mathrm{\&beta;}}_j^{\left(2\right)})=\frac{{\mathrm{\&beta;}}_j^{\left(1\right)}\&CenterDot;{\mathrm{\&beta;}}_j^{\left(2\right)}}{\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_j^{\left(1\right)}\&CenterDot;{\mathrm{\&beta;}}_j^{\left(2\right)}}$

Wherein, ${\mathrm{\&beta;}}_j^{\left(1\right)}=\{{\mathrm{\&beta;}}_1^{\left(1\right)},{\mathrm{\&beta;}}_2^{\left(1\right)}...{\mathrm{\&beta;}}_p^{\left(1\right)}\},$ Reflected of the contribution of the contained quantity of information of index itself, utilized aforesaid projection pursuit analytical model to try to achieve evaluation result. ${\mathrm{\&beta;}}_j^{\left(2\right)}=\{{\mathrm{\&beta;}}_1^{\left(2\right)},{\mathrm{\&beta;}}_2^{\left(2\right)},...{\mathrm{\&beta;}}_p^{\left(2\right)}\},$ Reflect the decision maker to the preference of each key element or attribute or the significance level of attribute itself, be the whole parameter of people's wage adjustment.

IV) interzone similarity evaluation.Be used to be specified to cultivated land district and comprehensive similarity and the similarity grade thereof of waiting to implement the area.Suppose that regional A comprises φ index, regional B comprises

Individual index, the coincident indicator of regional A and B have σ,

The similarity S between j the similar index wherein _jBy II) step draw.By the definite similarity degree S of index quantity _nExpression is by similarity S _jThe similarity S that determines _uExpression, β _jExpression is by III) complex weight of the j that obtains of a step similar unit, it had both considered decision maker's preference, had also taken into account the self information of index.Similarity S between area A and the B ^*(A, mathematical model B) is expressed as follows:

$S_u={\mathrm{\&beta;}}_1\&CenterDot;S_1+{\mathrm{\&beta;}}_2\&CenterDot;S_2+...{\mathrm{\&beta;}}_{\mathrm{\&sigma;}}\&CenterDot;S_{\mathrm{\&sigma;}}=\underset{j=1}{\overset{\mathrm{\&sigma;}}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_j\&CenterDot;S_j.$

In the step 5) of the inventive method, model selection is used for according to the ripe area of conservation tillage and waits to implement the similarity grade in area, when the similarity grade is I, II level, for waiting that implementing the area selects suitable conservation tillage pattern.The corresponding relation of similarity grade and similarity value following (as shown in table 2):

Table 2: the corresponding relation of similarity grade and similarity value

Wait to implement area and ripe regional pattern match more than the process, so just need not time consuming field experiment, the present invention is the conservation tillage pattern that the most suitable their actual conditions have been selected in area to be implemented.

Conservation tillage model selection model of the present invention has analytical approach advanced person, result advantage accurately.Wherein the definite of index finishes by on-site inspection and expert's informal discussion method; Projection pursuit technique is determined objective weight

Process in treat index with a certain discrimination, adopt diverse ways to carry out data normalization according to the characteristics of different indexs, when the optimization aim function, adopt based accelerating genetic algorithm and MATLAB programming based on real number.In the evaluation procedure of system similarity degree, adopt distinct methods to determine each index similarity according to the difference of different index properties.Complex weight β _jDerive from objective weight

With subjective weight

Utilize aforementioned projection pursuit technique to determine, Obtain after utilizing the comprehensive expert opinion of expert survey.

What protective farming mode selection method of the present invention can be different weathers, geography, production custom waits that implementing the area selects suitable conservation tillage pattern.

Claims (2)

1. protective farming mode selection method, it is characterized in that: it may further comprise the steps:

1) data aggregation: soil data, climatic data, geodata, the production custom data of collecting the ripe area of different mode; Collection waits to implement soil data, climatic data, geodata, the production custom data in area;

2) model selection is prepared: according to the data in described ripe area, determine the decision package and the index of each conservation tillage pattern, corresponding decision package of the classification in each ripe area wherein, described index derives from soil data, climatic data, geodata, the production custom index that has comparability in the ripe area;

3) make up the projection pursuit analytical model: detection, the enquiry data in ripe area are carried out standardization,, determine the objective weight of each index with projection pursuit technique according to described decision package and index U=1 wherein ..., n, p are the number of index, make up the projection pursuit analytical model and may further comprise the steps:

A) data normalization: with raw data matrix X ^*Be converted into standardization matrix X, x _IjBe the element among the standardization matrix X, represent j desired value of i decision package;

B) linear projection: establish vector of unit length a={a ₁, a ₂... a _pBe the one-dimensional linear projecting direction, then to project to the one dimension projection properties value on the projecting direction a be z to standardization matrix X _i

C) structure projection target function:

Q(a)＝S _Z·D _Z

Wherein, E (z) is sequence { z _i| i=1 ..., the mean value of n|}; If r _IkBe the distance between the projection properties value; N is the classification number in ripe area; J=1,2 ..., p; R is the density window width, and its span is:

r _Ik=| z _i-z _k|, i, k=1 ..., n;

D) optimize projection target function Q (a): the maximization of utilization objective function is optimized projecting direction a the projection target function, make projection target function Q (a) reach maximum value projecting direction be the best projection direction

MaxQ(a)＝S _z·D _z

E) objective weight is calculated: the objective weight of determining described each index

4) make up the similarity evaluation model: according to the data in each ripe area, parameter similarity complex weight β _jCalculate the similarity value of waiting to implement between area and each the ripe area afterwards, estimate similarity and the similarity situation waiting to implement between area and each the ripe area, make up the similarity evaluation model and may further comprise the steps:

I) index classification;

Ii) calculate the index similarity S wait to implement same index between regional A and ripe regional B (A, B), it may further comprise the steps:

1. when the value of two indexs all was determined value, the step of parameter similarity was as follows: when obtaining determined value behind the step a) data normalization in the index process step 3) of regional A and B, and index L (A), L (B) ∈ [0,1]; Similarity between the determined value can represent that distance is near more with the distance of point-to-point transmission, and similarity is high more, then has:

2. when the index value of regional A be a determined value, and the value of the same index of regional B is when being an interval value, the step of parameter similarity is as follows: L (A) ∈ [a ₁, a ₂], the index L of regional B (B) is through the interval that obtains after the standardization, L (B)=[b ₁, b ₂], a ₁＜a ₂, b ₁＜b ₂, and a ₁, a ₂, b ₁, b ₂∈ [0,1]; Similarity between determined value and the interval, relevant with the distance and the length of an interval degree of determined value and interval mid point respectively, the distance of determined value and interval mid point is more little, and the length of an interval degree is short more, and similarity is high more;

If when 3. the value of two indexs all was interval value, index calculation of similarity degree step was as follows: the similarity between two intervals, adopt interval level of coverage to represent, level of coverage is big more, and similarity is high more;

4. binary variable index similarity is calculated: the value of this index of binary variable index expression has only 0 or 1 two state, if regional A and regional B have several same binary variable index, then the similarity calculation procedure of all identical binary variable index is as follows: establish h and represent that the value of the same index of the index of regional A and regional B all is 1 index number, f represent the index value of regional A be 1 and the value of the identical index of regional B be 0 index number, s represent the index value of regional A be 0 and the value of the identical index of regional B be 1 index number, t represents that the value of the same index of the index of regional A and regional B all is 0 index number, d (A, B) distinctiveness ratio of the whole identical binary variable index of regional A of expression and regional B, be their Euclidean distance, then the similarity of whole identical binary variables is:

5.; If regional A and regional B have several nominal data indexs, then all the similarity calculation procedure of nominal data indexs is as follows: the nominal data index is the popularization of binary variable index, promptly in the index amount of state of value more than two; Suppose that g represents the index number that regional A is identical with the value of same nominal data index among the regional B, q represents the number of all nominal data indexs among regional A and the regional B, d (then the nominal data similarity is for A, the B) distinctiveness ratio of the nominal data index of regional A of expression and regional B:

If 6. two indexs all are qualitative index, then qualitative index similarity calculation procedure is as follows: the similarity of qualitative index, usually will bluring qualitatively, index is mapped to the numerical value description, comparison by numerical value, utilize the similarity formula of step two determined value indexs 1., thereby realize the qualitative of similarity to Quantitative yield;

Iii) index similarity complex weight is definite:

Wherein, described

Try to achieve by described projection pursuit analytical model;

The parameter that the behaviour wage adjustment is whole;

Iv) interzone similarity evaluation: regional A comprises φ index, and regional B comprises

Individual index, φ≤p,

The coincident indicator of area A and B has σ,

The similarity S between m the similar index wherein _mEqual by the ii) index similarity S of m index drawing of step (A, B), m=1 ... σ; β _mFor by ii i) the similarity complex weight coefficient of m index obtaining of step; Similarity S between area A and the B ^*(A, mathematical model B):

5) model selection:, select to wait to implement the regional pairing conservation tillage pattern of the highest described maturation of regional similarity, as the conservation tillage pattern of waiting to implement the area with described according to described similarity grade.

2. a kind of protective farming mode selection method as claimed in claim 1 is characterized in that: in the described step 1), the soil in described ripe area is data from existing experimental result and on-site inspection result; The described soil The data chemical gauging of waiting to implement the area; Described maturation is regional and wait that the climatic data of implementing the area obtains according to 10 years meteorological data weighted means; Described ripe area obtains with the The data field method of the geodata of waiting to implement the area, production custom.

Images