TuyIn tap Cong tnnh Nghien ciiu Cong nghe Thong tin va Truyin thong 2010
DlT BAO MlTC DO NHIEM RAY THEO THOHL GIAN
Vii Duy Linh, Le Quyet Thang, Huynh Xuan Hif p
Khoa Khoa Hgc Ty Nhien - Trucmg Dai hoc Cin tha, Khoa CNTT&TT - Tnrdng Dai hoc Cin Tha {vdlinh, Iqthang, hxhiep)@ctu.edu.vn
Tom tit. Su dung cong cu cong nghe thong tin trong mo hinh hoa cong tac du bao dich hai, nhat la doi voi sy xuit hien ciia riy nau la mpt van de cap thiet hien nay tai khu vyc phia Nam trong nganh bao ve thyc vat.
Viec xac dinh dugc cac miic do nhiem ray d ngay thdi diem hien tai va dy bao dugc cac muc dp nhiem ray d tucmg iai gan se giiip cho cac chuyen gia cflng nhu ngudi nong dan chii dong su dung cac bien phap phong trir ray thich hgp de bao ve cay Ilia. Moi quan he giira viec xuat hien ray nau va cac tac dpng cua moi trudng la mpt tiep can dugc nhieu chuyen gia trong nganh bao ve thyc vat ap dung \h da dugc cong nhan la nhiing tiin bo khoa hpc ky thuat hien nay. Trong bai bao nay, chung toi de xuat hudng tiep can mdi la sti dung mang Bayes kit hgp vdi xich Markov nham danh gia sy tac dpng qua lai cua cac yeu td moi trudng theo thdi gian trong viec dy bao. Kit qua thu dugc tir mo hinh nay se dong vai tro ho trg quyet dinh cho cac chuyen gia bao ve thyc vat tai Trung tam Bao ve Thyc vat Phi'a Nam trong qua tnnh du bao cac miic dp nhiem ray theo thdi gian.
Tu khoa: Nhiem ray nau, xich Markov, mang Bayes, dy bao theo thdi gian.
TEMPORAL FORECAST FOR EPIDEMIC DISEASE LEVELS OF BROWN PLANT HOPER
Abstract Using tools of information technology for modeling the forecast of epidemic diseases, especially the appearance of brown plant hopers (BPHs) as well as the variety of BPHs density is very important nowadays. Determining epidemic disease levels in present time and forecasting how epidemic disease levels will happen in the fiiture is a practical and useful thing. This will be helpful for agricultural specialists and farmers to know in advance to prevent and protect rice fields from BPHs actively and effectively. The relation between BPHs' appearance and environmental effects is an advanced science and technology approach which has been applied by specialists. In this article, we suggest a new method for evaluating the interaction of environmental factors that change temporally. This method is to combine Bayesian networks and Markov chain for forecasting BPH epidemic disease levels.
Keywords: BPH epidemic disease, Markov chain, Bayesian network, temporal forecast.
Tuyen tap Cong trinh Nghien cim Cong nghf Thong tin va Tniyin thong 2010
Dir BAO MirC DO NHIEM RAY THEO T H 6 I GIAN
Vli Duy Linh
Khoa Khoa hgc Ty nhien, Trudng Dai hgc Cin Tho c i n Tho, Vif t Nam
vdhnh @ctu.edu.vn
Le Quyet Thing
Khoa CNTT&TT, Trudng Dai hgc Cin Tho c i n Tho, Vif t Nam
[email protected] Huynh Xuan Hiep
Khoa CNTT&TT, Trudng Dai hgc Cin Tho c i n Tho, Viet Nam
hxhiep @ctu.edu. vn Tom tat — Sil' dung cong cu cong nghe thdng tin trong
mo hinh hoa cong tac dy bao djch hai, nhat la doi vdi sir xuat hifn ciia ray nau la mot vin de cap thiit hien nay tai khu vuc phia Nam trong nganh bao ve thyc vat.
Vifc xac djnh daqtc cac mu-c dp nhilm ray if ngay thoi diem hifn tai va du- bao durffc cac muc do nhilm ray d tuong lai gan se giiip cho cac chuyen gia cung nhir nguM nong dan chii dfng su dung cac bien phap phong trir ray thich hgp de bao ve cay lua. Moi quan he giira vifc xuat hien ray nau va cac tac dpng cua moi trudng la mot tiep can duo'c nhieu chuyen gia trong nganh bao ve thyc v3t ^p dung va da duffc cdng nhan la nhirng tien bo khoa hoc k>^ thuat hifn nay. Trong bai bao nay, chting toi de xuat hudng tiip c|n mdi la sir dung mang Bayes ket hgp vdi xich Markov nhSm danh gia sy tac dpng qua lai ciia cac yeu to moi truong theo thoi gian trong vifc dy bao. Ket qua thu duyc tir mo hinh nay se ddng vai tro hS try quylt dinh cho cac chuyen gia bao vf thyc v3t tai Trung tam Bao vf Thyc v3t Phia Nam trong qua trinh dy bao cac mii-c dp nhiem ray theo thai gian.
Tie khoa — Nhiem ray nau, xich Markov, mang Bayes, dubdo theo thai gian.
I. G I 6 I THIEU
Trong ITnh vuc trdng va tham canh cay lua, ray nSu 1^ lo^i cdn trung gSy hai Idn nhat va phd bien nhat hifn nay [2] [7]. Ngoai vifc chich hut tryc tiip v&o cay Kia, nd con la ddi tugng chfnh truyen bfnh vang liin, lim xoan la, nlu khdng phong tri kjp thdi se gay thift hai rat Idn den nSng suat cua ba con ndng dan [14].
Vdi d^c difm gay hai cua ray nau nhu vay, vifc xdc djnh dugc miic do nhilm riy d thdi diem hifn tai Cling nhu dy bao dugc miic do nhiem se dien tien nhu the nko d mdt thdi diem trong tuong lai de c6 hifn phip phdng chdng hifu qua la mdt nhu cau thue tien cap bdch can duoc giai quyet. Do vay, dl cd thl giiip hd trg cho cac cSp chinh quyen cd dugc nhung ca sd dy bk> chinh xac trong cong tac chi dao va phdi hgp h^nh dgng de phdng chdng ray tren difn rdng, md hinh dy bdo miic dg nhiem ray da dugc tao ra de phuc
vy cho vifc dieu hanh va hd trg cho ba con ndng dan d 22 tinh thanh phi'a Nam [4] [7].
Mdt sd md hinh dy bao da dugc ling dyng tren thl gidi nhu: van de vl quan ly ngudn tai nguyen thien nhien va hf sinh thai [8] [15], bao vf giu frang d Bic eye [1], danh gia mdi trudng sdng thich hgp cho tuin lgc rimg d Bic Columbia [11],... da su dung tri thiic chuyen gia vao frong mang Bayes [3] dl dy bao. Tuy nhien, cac md hinh nay chua phan anh dugc sy ty bien ddi ben trong cac yeu td theo thdi gian do sy tac ddng cua cac mdi trudng xung quanh dya tren cac nguyen tic cua xich Markov [6].
Trong bai viet nay, chiing tdi gidi thifu mgt tiip can mdi nhim ho trg vifc ra quylt djnh dy bdo miic dg nhilm riy bing each su dung cdc ky thuat mang Bayes va xich Markov. Cdc yeu to nglu nhien tuomg tac theo luat nhan qua 1^ cdc yeu td dau vao ciia mang Bayes, cdc yeu td nay se bien ddi theo thdi gian vl dugc xdc djnh bdi xich Markov. Do vay, cdc yeu td dau vao tuong tdc theo luat nhan qua va thay ddi theo thdi gian nay chinh la cdc yeu t l dau vao cho md hinh dy bdo theo thdi gian.
Bai viet nay dugc td chiic thanh nam phin. Phin thii nhat gidi thifu chung ve hudng tiep can cua cdng tdc dy bao nhiem ray tren co so su dyng tiep can mang Bayes va xich Markov. Phan thii hai trinh bay so luge ve hai ky thuat dugc sii dyng trong md hinh.
Phan thu ba xay dyng md hinh dy bdo nhiem ray trong dd vifc tich hgp ky thuat xich Markov vko mang Bayes de hinh thanh vifc dy bdo nhilm ray theo thdi gian. Phan thii tu trinh bay vifc thu nghifm md hinh dy bdo nhiem ray. Tdm tit mgt sd ket qua quan trgng va mdt sd hudng phdt trien dugc neu ra d phan cudi ciing.
H. MANG BAYES VA XICH MARKOV A. Mgng Bayes
Mang Bayes [3] [5] [8] [16] la mdt dd thi cd hudng khdng cd chu trinh (Directed Acyclic Graph, DAG) G = (V, E) vdi cdc nut V = {v,,..., v„) vd tap cac cung E, mdt tap cdc bien rdi rac ngau nhien, X, dugc dai difn bdi cac mit trong G (xem Hinh 9), vl mgt tap cdc phan phdi xdc suat cd dieu kifn
TuyIn tap Cong trinh Nghien ciiu Cdng nghe Thdng tin va Truyin thong 2010
(conditional probability distribution) P, ling vdi mdi biin ngiu nhien X , G X, nd chiia dyng phan phdi xac suat cd dieu kifn cua P(Xv I Xparcntgv))- Vdi parent(v) la cac niit cha cua v. Cac phan phdi xac suat cd dieu kifn dugc td chirc thanh dang bang va dugc ggi la CPT: Conditional probabihty table (xem Bang IX).
B. Xich Markov
Xich Markov [6] la mgt qua tnnh ngau nhien dac bift khdng phu thugc vao qua khii ma chi phu thugc vao hifn tai. Xet mdt xich Markov {X,},gx va thdi diem ban dau t = 0, he thdng cd trang thai Xo = CQ.
Yeu cau cua bai toan nay la dy bao cdc kha nang xay ra cua he thdng sau t don vi thdi gian dudi dang mdt phan phdi xdc suat cua bien X, cd dieu kifn Xo= eo da xay ra.
Bai toan dy bao la tinh xdc suit de hf thdng dat trang thai x tai thdi diem t khi bilt tai thdi dilm t = 0, Xo = Co, hay tinh P(X,= x I Xo = eo). D I giai bai todn dy bao trong trudng hgp thdi gian rdi rac T = {0, 1, 2 , . . . } , ta se dp dung cdng thiic xac suat toan phan de tinh xdc suat tai thdi diem t thdng qua xac suit tai thdi diemt-1:
P(,X, =x\ X„ =e„) = Y,P(X,_, =k\X„ =e„)P(X, =x\X_, =k,X„ =e„)
kEE
N I U {X,} la xich Markov, do tinh khdng phu thugc vao qud khii, nen ta cd:
P(X, = XI X,_, =k,X,= e,) = P(X, = XI X,_, = k) P{X,=x\X,=e„) = Y,P(X„,=k\X„=e,)P(X,=x\X,^=k)
Ky hifu V, la vec-to chuyen vi cua vec-to v , . Phuang trinh Chapman-Kolmogorov dudi dang ma tran se nhu sau:
V,=V,_,P = V,_^P^=-- = V,P' (1) Trong dd: P la ma tran xac suit chuyin trang thai, V 0 la vec-to trang thai ban d i u cua hf thing va dugc xdc dmh dya vao trang thdi ban diu tai thdi diem t = 0:
v,(x) = P(X,=x\X,=e,) = \\''^^'
\y,x^e^
Nhu vay, vec-to Vj, cd mgt phin tu bing 1 (tiiang ling vdi tiang thdi ban diu eo), cdn tit ca phin tii khde bang 0. Vec-to V Q cd dang:
1 2 ••• e„ e„ + l ... n
^0 ="• 0 0 0 1 G O O ]
Xich Markov cd thl du bdo ttong t don vi thdi gian va d mdt thdi dilm bit ky ttong nrong lai nlu ton tai qud ttinh dimg (stationary process) ttong hf thdng dy bdo. V l mat ung dung, d l kilm tta qud tnnh dimg, cdc ttang thai ciia mdi nut nhap ttong mang Bayes phai cd mdi lien hf la lien thdng manh (strongly connection) [6].
HI. M O HINH D V BAG MlTC DO NHIEM RAY
De cd thl tinh todn dugc miic do nhilm ray d thdi diem hifn tai ciing nhu dy bdo dugc miic do nhifm ray se diin ra nhu t h l nao ttong tuomg lai la mdt qud tnnh thue hifn p h u c tap, ket qua cua nd phu thugc vao nhieu yeu td khac nhau ttong mdi trudng sdng ty nhien ciia cay lua nhu [2] [7] [14]: Do tudi ray, gidng Ilia, phan bdn... Cac yeu td nay cd tinh riglu nhien, tuang tac nhau theo luat nhan qua va d dieu kifn suy dien khdng chac chan. Vdi nhirng dac diem nhu vay, chiing tdi da sii dung mang Bayes va xich Markov lam nen tang cho md hinh du bao.
A. Md hinh ddnh gid mite do nhiem rdy a hien tgi Md hinh dy bao nhiem ray la su tuong tdc giiia cdc ddi tugng/tdc nhan ttong mdi trudng sdng cua cay lua. Cdc yeu td quan ttgng cung vdi cdc ttang thdi tdc ddng tryc tiep vao mdi trudng sdng cua cay Ma ciing nhu quyet dinh den miic do nhilm ray dugc lift ke ttong Bang I.
BANG I. CAC TAC NHAN CHINH TRONG Mdi TRUONG
CAY LOA.
Ten tac nhan (doi tirong) Giong lua
Tinh khang ray (ciia Ilia) Goc thuoc trii sau Ng&y xjt thuoc sau s^
Lirgng phan &a.Ta tiSu chuin
Thoi diem bon phan Mat do sa Thien dich
M$t so ray ngo^i dong Dp tuoi ray
Giai doan sinh truong lua
Trang thai (ti.uoc tinh) tac nhan IR64 Jasmine85,MTL499,...
Khing, trung binh, nhiem Clorpyrifos, cypermethrin, Vhis Duoi 40 ngky, tren 40 ng&y Nho hon 90, tir 90 den 100, lorn hon 100 (kg/ha)
Dung, sai
Duoi 120, Uen 120 (kg/ha) It, nhieu
Duoi 5000, tir 5000 den 10000, tren 10000 (con/m^)
Tning, tudi 1 den 3, tuoi 4 d£n 5, tudi trudng thinh
Ma, de nhanh, dong tro, chin
Cac tdc nhan d tten se sinh ra cac ylu td tmng gian ttong md hinh du bao theo luSt nhan qua. Trong md hlnh t o n | quat, cac ylu td tac ddng tmc tiip len miic dg nhiem ray dugc xac djnh la cac tac nhan:
"Mat sd ray ngoai ddng", "Mat sd r i y biin ddng" vk
"Miic gay hai cua r i y " Cac ylu td nay dugc kit hgp vdi nhau thdng qua cau tnic va bang phan phdi xdc suat cd dieu kifn - CPTs, ma chung tdi ggi la co so tii thiic, cua mang Bayes d l suy ra miic do nhilm riy d thdi diem hifn tai (xem ITinh 1).
Matsoiayi^oaidSng Mat so riy bien d6i% Miic g^ hai ciia i^
Tmh toan va si^ dien dua tren cff sor tti 4iic CPTs
T
Miicdpnhianray
Hinh 1. Md hlnh du bio d thdi dilm hifn t^i
Tuyen tap Cong trinh Nghien ciiu Cong nghf Thong tin va Truyin thong 2010
Mat sd rdy ngodi ddng: Y I U td nay dugc xdc dinh thdng qua viec guan ttic, thu thtp sd lifu dinh ky 7 ngay mdt lin, bat diu tit liic lua d do 7-10 njgay hidi va cho din liia chin sap [2] [7]. Thdng tin ve mat sd riy ngoai ddng dugc tien hanh theo khay (ddi vdi lua cay) dl dilu tta tirng khdm (danh) mdt, va diing khung (ddi vdi lua sa va ma) de dim tiirc tiip sd riy nau Cling nhu phan do tudi. Trong thue tl, ngudi ta thudng phan ra ba khoang vl mat sd ray dd la: it hon 5000, tir 5900 din 10000, va tten 100000 (con/m').
Mgt so rdy bien ddng.- Viec xdc djnh cdc gia ttj cho yeu td mat sd riy biin ddn^ cin dugc ti'nh toan qua cdc gid tti cho mdt sd ylu td ban diu. Cdc ylu tl dd, bao gdm mdi quan hf nhSn qua va bang gid tti phan phdi xdc suit se dugc ttinh bay ngin ggn nhu sau:
Giong Ilia Tinh khang ray Hinh 2. Mdi nhan qua ciia yeu td 'Ti'nh khang rSy"
Ve mat ket qua sinh ra dya tten luat nhan qua, chdng ta cd the xem gid tti cac ttang thdi cua ylu td 'Tinh khang riy" dudi tdc nhan "Gidng liia" nhu la mdt ham cd tham sd Fi(X), vdi X la "Gidng lua". Cdc gid tti xac suat cua ham Fi dugc tinh todn vd trich lgc mdt sd hang nhu Bang II.
BANG n . X IR64 JasmineSS MTL499 OM4900 OMCS2000 VD20
CAC GIA TRI CUA HAM F|(X) F,(X)
khdng 0,00 0,00 0,00 0,00 0,00 0,00
trung binh 1,00 0,00 1,00 1,00 1,00 0,00
nhiem 0,00 1,00 0,00 0,00 0,00 1,00
Tinh khang ray
• '
MStdfs?
' ' Ke hoach gieo s$ giong
Hlnh 3. Mdi nhan qua ciia yeu td "Ke hoach gieo s^ gidng"
Cung theo luat nhan qua, "Kf hoach gieo sa gidng" duoc thl hi6n nhu la ham hai tham sd F2(X, Y) vdi X ia "Tinh ididng riy" va Y la "Mat do sa"
Cdc gid tri xdc suit cua ham FT dugc tinh d Bang HI.
BANG n i . X khing khdng trungblnh truii^blnh nhilm nhiem
CAC GIA TRI CUA HAM FiiX, Y) Y
dudi 120 tren 1:0 dudi 120 tren 120 dudi 120 trSnl20
F2(X,Y) hvply 1,00 0,50 0,80 0,40 0,40 0,00
khdng 0,00 0,50 0,20 0,60 0,60 1,00
Gdc thuoc tni sau Ngay xit thuoc sau s?i
Ke hoach phun xit thuoc
Hinh 4. Mdi nhSn qua ciia ydu td "K6 ho^ch phun xjt thude"
Tuomg ty, ylu td " K I hoach phun xit tiiude" la F3(X, Y) vdi X la "Gdc thude trii sau" va Y la "Ngay xjt thude sau sa" Cdc gid tti cua hdm F3 dugc cho nhu Bang IV.
BANGrV. CACGIATR!CUAHAMF3(X,Y) X
khac khac clorpyrifos clorpyrifos cypermethrin cypermethrin
Y tren 40 ngsly duoi 40 ngay tren 40 ngay dudi 40 ngay tren 40 ngay dudi 40 ng^y
F,(X,Y) hfply 0,90 0,20 0,30 0,10 0,50 0,15
khong 0,10 0,80 0,70 0,90 0,50 1,00
Luang phan dam tieu chuan Thai diem bon
Ke hoach bon phan
Hinh 5. Mdi nhan qua ciia yeu td "Ke hoach bdn phSn".
Tuong ty, " K I hoach bdn phan" la ham F4(X, Y) vdi X la "Lugng phan dam tieu chuin" va Y la 'Thdi diem bdn phan". Cdc gid tti ciia ham F4 dugc cho nhu Bang V.
BANG V.
X nhd hom 90 nhd hom 90 tur 90-100 tir 90-100 idn hon 100 Idn hem 100
CAC GIA TRI CUA HAM F4(X, Y) Y
dung sai Uen 40 npiy dudi 40 ngay tren 40 ngSy dudi 40 ngay
F4(X, Y) h(tply 0,90 0,50 0,30 0,10 0,50 0,15
khong 0,10 0,50 0.70 0,90 0,50 1,00
Sau khi da xac dinh dugc cdc yeu td: "Ke hoach bdn phan", "Ke hoach xjt thude sau", "Ke hoach gieo sa gidng" va 'Thien djch", ySu td "Mat sd ray bien ddng" cd mdi quan hf nhan qua nhu Hinh 6.
Kehosch bon phan
I
Mat so ray bien d$ngKe hosch xit thuoc sau
Ke hoach gieo sa gilng
I I
Thiln djch Hinh 6. Mdi nhSn qua cua ylu td "M^t sd ray biin dpng"
Tuong tu, "Mat sd ray bien ddng" Id mgt hdm cd bin tham sd F5(X, Y, Z, T) vdi X: "Kf hoach gieo sa gidng", Y: " K I hoach xit tiiude sau", Z: " K I ho^ch bdn phan", va T: 'Thien djch" Gid ttj cua hdm F5
TuyIn tap Cong trinh Nghien ciiu Cong nghe Thdng tin va Truyin thong 2010
dugc tinh va ttich lgc mdt sd hang nhu ttong Bang VI.
BANG VI. CAC GIA TRI CUA HAM FsCX, Y, Z, T) X
hop ly hgrp ly khong khdng khdng khdng
Y hop ly khdng hgp ly hgp ly khdng khdng
Z hgp ly khdng hgp ly hgp ly khdng khdng
T nhieu It nhieu it nhieu ft
FsCX, Y, Z, T) giam
1,00 0,30 0,55 0,75 0,22 0,00
tang 0,00 0,70 0,45 0,25 0,78 1,00
Mii-c gdy hgi ciia rdy: dugc xdc dinh bdi hai tdc nhan theo mdi quan hf nhan qua giua "Do tudi ray"
va "Giai doan sinh trudng lua" nhu Hinh 7.
Miic gay hai cua r a y
D Q tuoi ray
T
Giai d o a n sinh t r u d n g Hinh 7. Mdi nhan qua ciia yeu td "Miic gay hai cua ray"
Tuomg ty, yeu td "Miic do gay hai cua riy" la mdt ham F6(X, Y) vdi X la "Do tudi riy" cd bin ttang thai: tning, tudi 1 den 3, tudi 4 den tudi 5, tudi trudng thanh; cdn Y la "Giai doan sinh trudng liia" cflng cd bdn ttang thai: ma, de nhanh, ddng ttd, chin. Cac gid trj xdc suit cua ham Fg dugc cho ttong Bang VII.
BANGVn. CACGIATRICUAHAMF6(X,Y) X
trimg trimg trimg tuoi1-3 tuoi1-3 tudi1-3 tudil-3 tudi4-5 tudi4-5 tudi-tt
Y ma de nhdnh chin ma de nhanh ddng trd chin ma de nhanh chin
F6(X,Y) nhe
1,00 1,00 1,00 0,05 0,00 0,00 1,00 0,10 0,05 0,95
trung binh 0,00 0,00 0,00 0,10 0,10 0,20 0,00 0,70 0,80 0,05
ngng 0,00 0,00 0.00 0,85 0,90 0,80 0.00 0,20 0,15 0,00
Din day ylu td "Miic do nhilm riy" da cd thl xdc dinh dugc thdng qua mdt hain co ba tham sd la F7(X, Y, Z) vdi X bilu thi cho ylu td "Mat sd riy ngodi ddng", Y cho ylu td "Mat sd riy biin dgng" hay ham Fj, va Z cho ylu td "Miic gay hai ciia riy" hay ham Fe- Mien xdc dinh cua chun§ dugc cho nhu sau: X e [Dudi 5000, tir 5000 den 10000, tten 10000 (con/m-)], Y e [giam, tang], Z e [nhe, tmng binh, nang], F7 6 [nhe, trung binh, nang]. Gid tti cua ham F7 dugc tinh nhu Bang VIII.
BANG VIII. CAC GIA TRI CUA HAM Pj(X. Y, Z) X
difdiSOOO
Y
giam
z
nh^
F7(X, Y, Z) nhf
0,95
trung binh 0,05
ngng 0,00
X
dir6i 5000 duoi 5000 dual 5000 dirori 5000 dirdi 5000 tir 5000-10000 tir 5000-10000 tir 5000-10000 tir 5000-10000 tir 5000-10000 tir 5000-10000 u-en 10000 U^en 10000 tren 10000 tren 10000 u-en 10000 U^en 10000
Y
giam giam t5ng tang tSng giam giam giam tang tSng tSng giam giam giam t^g tSng ting
z
tmng binh ning nh?
tnmg binh nSng nh^
trung binh ning nhf trung binh ning nhf trung binh ning nhf trung binh ning
F7(X, Y, Z) nhf
0,70 0,40 0,30 0,22 0,13 0,50 0,37 0,21 0,16 0,12 0,07 0,00 0,00 0,00 0,00 0,00 0,00
trung binh 0,30 0,55 0,60 0,68 0,73 0,50 0,63 0,74 0,74 0,78 0,79 0,50 0,50 0,48 0,45 0,45 0,05
tfng 0,00 0,05 0,10 0,10 0,14 0,00 0,00 0,05 0,10 0,10 0,14 0,50 0,50 0,52 0,55 0,55 0,95
Giong lua
I
Tinh khang riy lugt dp sa Goc thuoc sau Ke ho^ch gieo s^ giong
Ngay xit thuoc sau sti
z
Luong phan dam tieu chuan Mat so ray bien dpng
Ke hoach phun xjt thuoc
~7
Thoi dilm bon phan
Ke hoach b6n phan
Thien dich
Su tuong t k theo luat nh§n qui giUa cic yeu 16 anh hudng toi mire do nhiem ray
2:
Dp tuoi ray
Miic ^ y hai ciia ray I i 0 so riy ngoai dong
X
Giai doan sinh tnrdng lua
Hinh 8. Cac yeu td tuomg tac theo luat nhan qua trong md hinh du bSo.
Dya vao ciu ttiic tuomg tdc d Hinh 8, chiing ta nhan thay cdc ylu td da dugc khao sat d phin tten la phil hgp vdi ban chit va chiic nang cua niang Bayes:
Suy dien dya yao luat nhan qua dudi dilu kifn suy diin khdng chac chan. Do vay ky thuat mang Bayes da dugc chung tdi chgn lua de thyc hi|n vifc cdi dat ttong hf thdng du bao miic do nhilm ray. Dilm thich hgp nua, dd la so dd cac ylu td tuong tdc d hinh ttSn cd dang phu hgp vdi cau ttuc mang da kit ndi (Multiply-connected network) va cdc gid tti cua cdc ham Fj cd the dl dang cai dat vao ttong cac bang CPTs - CO sd tri thiic- tuong ling cho mdi niit ttong mang. Ddng thdi, dua vdo ky thuat suy diin chinh xdc tten cau true mang da kit ndi. Junction tree, chiing ta CO the suy difn dugc ket qua vl miic do nhifm ray cin tim d thdi dilm hien taidan^ xet [3] [5] [8] [16].
Sau khi da thiet ke xong phan md hinh, vifc tiip din la cai dat he thdng dy bdo d thdi dilm hifn tai thdng qua cdng cu mang Bayes nhu sau:
Tuyen tap Cong trinh Nghien ciiu Cdng nghf Thong tin va Truygn thong 2010
Phin ciu tnic mang Bayes (xem Hinh 9): Trong phan nay, cd hai loai niit: Niit tdc ddng dugc ggi la niit khdng cd cha hay mit nhap, va nut bi anh hudng theo luat nhan qua dugc ggi la mit cd cha. Trong hf thing ndy, cd 10 mit nhap do la: Gidng lua, mat do sa, gdc thude trir sau, ngay xit thude sau sa, lugng phan dam tieu chuan, thdi diem bdn phan, do tudi riy, giai doan sinh trudng liia, thien dich va mat sd riy ngoai ddng. Cdn 7 mit cd cha la: Ti'nh khdng riy, ke hoach gieo sa gidn^, kl hoach xit thude sau, kl hoach bdn phan, mat sd riy bien ddng, miic gay hai cua riy, va nut mirc do nhiem ray.
Hinh 9. CSu tnic m^g Bayes de du bio miic dp nhilm ray.
_ Phan thii hai la cdc bang co sd tti thiic dugc bieu dien thdng qua cdc bang CPTs. Ddi vdi mdi niit nhap se CO mgt bang CPT ma gia tti dugc tinh bdng vifc thdng kS sd lifu dugc thu thap hang tuan tir ngoai ddng rugng. Ddi vdi cdc mit cd cha, cac gia tti cua chdng dugc tinh dya vao tti thiic chuyen gia. Trong phan ndy, chiing tdi chi minh hga mdt bang CPT (chinh la ham F7) cho mit "Miic do nhiem ray", MDNR nhu Bang IX.
BANG IX.
M | t sd riiy ngoai ding
dmSriSOOO dii6i 5000 du6i5000 duAi 5000 dudi 5000 duAiSOOO tir 5000-10000 tit 5000-10000 tir 5000-10000 tir 5000-10000 Iir5U00-10000
CPT CUA NUT "Mirc DO NHIEM RAY"
MSt so ray bien d9ng giam giam giam ting ting ting giam giam giam ting ling
Mirc gay hai cua ray
nhf tmng binh ning nhf trung binh ning nhf trung binh ning nhf trung binh
nhe
0,95 0,70 0,40 0,30 0,22 0,13 0,50 0,37 0,21 0,16 0,12
P(MDNR trung
binh
0,05 0,30 0,55 0,60 0,68 0.73 0.50 0,63 0,74 0.74 0,78
=
n i n g
0,00 0,00 0,05 0,10 0,10 0,14 0,00 0,00 0,05 0,10 0,10
Mat so ray ngoai dong
tir 5000-10000 tren 10000 Uen 10000 uen 10000 uen 10000 tren 10000 tren 10000
Mat I so \
ray bien dong ting giam giam giam ting ting ting
Mirc gay hai ciia ray
ning nhe trung binh nang nhe trung binh ning
nhe
0,07 0,00 0,00 0,00 0,00 0,00 0,00
P(MDNR trung
binh
0,79 0,50 0,50 0,48 0,45 0,45 0,05
=
nang
0,14 0,50 0,50 0,52 0,55 0,55 0,95
B. Stf bien ddi gid tri cdc yen td ddu vdo theo thdi gian
Nhu da trinh bay d phan tten, hifn tai, md hinh mdi chi cd kha nang suy dien ket qua ve mirc do nhiem ray d tai mdt thdi diem. Do vay, vin de dat ra la he thdng dy bao cua chiing ta se nhu the nao neu mdi trudng xung quanh nd se bien ddi theo thdi gian [9] [10] [12] [13] d ttong Urong lai. Dl md hinh ttd thdnh mdt cdng cu dy bdo hiiu ich, no can dugc md rdng de cho phep sir dung ttong hf thdng thyc, tiic la anh hudng bdi yeu td thdi gian. Phuang phap dugc chung tdi chgn la sii dyng xich Markov de tien hdnh xu ly cdc gid ttj dy bdo mdi ttong tuong lai. Dieu nay ddng nghla vdi vifc dp xich Markov vdo mdi mit nhap (niit khdng cd cha) ciia mang Bayes dl giup cho he thdng cd the ty dy bdo dugc cdc gid ttj mdi. Tiip den, cac gid tti mdi nay se dugc ty ddng nhap ttd vao hf thdng de tiep tuc suy dien ra ket qua mdi vd dugc ggi la ket qua dy bdo ttong tuong lai. Ket qua la chiing ta xay dyng dugc md hinh dy bdo muc do nhiem ray theo thdi gian.
Trong ind hinh dy bao chiing tdi da kiem chiing dugc cdc yeu td dau vao deu Id nhirng qud ttinh dvrnjg.
VI dy so dd chuyen ttang thdi cua yeu td "Mat sd ray ngodi ddng" la mdt qua ttinh dvmg tdn tai ttong sudt hf thdng thyc nhu Hinh 10.
10.8
Hinh 10. Mat sd ray ngoki ddng dat trang thdi dimg Ma ttan chuyen ttang thai cho ndt "Mat sd ray ngoai ddng" dugc xdc djnh nhu cdc gid tti gia dinh sau:
0,80 0,09 0,11"
P= 0,34 0,62 0,04 0,30 0,41 0,29
De md hinh cd thl dy bao kha nang nhilm ray ttong tuomg lai, bing vifc thdng qua co sd ky thuat xich Markov chiing ta cd the dy bdo dugc gid ttj xdc
TuyIn tap Cong trinh Nghien ciiu Cdng nghf Thong tin va Truyin thdng 2010
suit mdi cua cdc ttang thai ttong mdi mit diu vao (mit khdng cd cha) ma chiing se thay ddi sau t don vi thdi gian (ngay) d tuomg lai. Tiep den, cac gia tti xdc suat mdi nay se dugc tu ddng nhap vao lgi chfnh cdc nut nay cua md hinh mang dl nd suy difn va tta ve ket qua dy bdo mdi.
Md hinh dy bdo theo thdi gian - sy tich hgp xich Markov vao mang Bayes dugc tiinh bay nhu ttong Hinh 11.
. C!Hi»idimhi8ntai ., ^^Cllfaiw dilm tuong la^.
Xich Markov - chuyen d5i gia tri cic trang
thai theo thai gian
Ccr so tn thiic CPTs:
,cic bang phan ph^i xic suat CO diSu kiSn
Ket qua hien tjii Kit qua dir bio
Ky hieu duoc sit dyng trong so do:
• buoc tiep theo « • gin kit, • • tie d$ng nhan qui Hinh 11. Md hinh mang Bayes (dg: bSo) theo thdi gian.
Giai thuat dy bdo cac gid tti ttang thai cua cdc yeu td dau vao ttong md hinh dy bao theo thdi gian tten ca sd ky thuat Markov:
for niiti c {tap cdc niit nhdp} do begin
read ma trdn xdc sudt P cda nUtj read vec-ta trgng thdi ban ddu Vo
for t =1 -^d(sd ngdy cdn du bdo, d< 90) do Compute vec-to a ngdy thu t, v,
theo cdng thirc (I)
update gid trj v, vdo bdng CPT ciia ndt, end
IV. KIEM THU* VA DANH GIA MO HINH Dir BAO
Vifc kiem thir md hinh dy bdo miic do nhilm riy se gua hai budc, do la kha nang suy diin miic do nhiem ray d thdi diem hifn tai va du bdo miic do nhilm ray d thdi dilm t ngay d tuong lai. Thdi gian dy bdo dugc chgn dl minh hga cho vifc kilm thii nay la t = 7 ngay.
A. Muc do nhiem rdy a thdi diem hien tgi
Gia su d thdi diem hifn tai (xem Hinh 11), cac yeu td dau vao dugc nhap vao cdc gid tti da dugc thdng ke tu ddng rugng. Vi du diin hinh nhu ylu to
"Mat sd riy d ngoai ddng" cd cdc gid ttj tuong ung cho cdc trang thdi cua nd la: dudi 5(X)0 con/m' chilm
85,48%, tir 5000 din 10000 con/m' chilm 4,5%, tten 10000 con/m-chilm 10,02%.
22.01 AlMril ISA] dGnglr^
K A I d*)
Hinh 12. Du bao d thdi di€m hi?n tai ting vdi cSc sd li?u vira thu thap.
Nhdn xet: Vdi cac niit diu vao dugc cap nhat cdc gia ttj xac suit mdi dugjc dilu tta thyc te tu ddng rugng nhu Hinh 12, he thdng se tinh todn cho cdc ylu td "Mat sd riy biin ddng" la tang = 55,82% va giam
= 44,18%; YIU td "Miic gay hai cua riy" la nhe chilm 61,62%, trung binh = 23,60 va 14,78 la iiang.
Hai ylu td nay cung vdi ylu td diu vao "Mat sd ray ngodi ddng", md hinh se dya vao cac bang CTPs - ca sd tti thiic, dl suy diin ra miic do nhilm riy d thdi dilm hifn tai la: nhe = 65,15%, trung binh = 26,93%, nang 7,92%. Nhu vay so vdi thyc tl, kit qua suy diin cua md hinh la hgp ly [2] [7].
B. Dif bdo mite do nhiem rdy d tuang lai sau t=7 ngdy
D I dl ttinh bay, chung tdi gia su rang, d hifn tai tit ca cdc nut nhap cd ttang thai thii hai la bang chiing (evidence), kit qua suy diin tim thay miic do nhiem riy d thdi dilm hifn tai dugc ghi nhan la: nhe = 8,77%, ttung binh = 78,39% va nang = 12,83% nhu Hinh 13.
Tuyen tap Cdng trinh Nghien ciiu Cdng nghf Thdng tin va Truyin diong 2010
Hinh 13. Du b5o d thdi diSm hi?n t^i vdi c4c bing chirng dugc xic l|lp.
Sau khi da xdc dinh dugc miic dg nhilm riy d thdi diem hifn tai, vdi mong mudn bilt dugc tinh hinh nhilm bfnh se nhu the nao ttong tucmg lai, chiing ta se cho hf thdng du bao. Vl mat ly thuyit, md hinh cd the dy bdo d nhieu thdi dilm ichdc nhau d mong lai (t <= 90 ngay). Tuy nhien, ttong thue tl chi cin biet trudc khoang 7 ngay la du thai gid dl ba con ndng dan phdng ngira giiip han che sy tac hai cua riy khi Ilia bi nhiem.
De thyc hifn dugc dilu nay, budc diu tien, md hinh se su dung kha nang dy bdo cua xich Markov dl dy bdo cdc gia tti mdi d tuong lai (xem Hinh 11) cd thl thay doi theo thdi gian sau t don vi thdi gian, nhu minh hoa d Hinh 14.
M|t so ray ngoal iJdng (con /rri2^
Mat so r9y ngoa dong (con /mC?
^^xfch Marlcov
$t sd lili ngoii a6ng (con / n ^
0.00 dua 5000 I tir 5000-10000
0.00 tr&l 10000
Mat se riy nqoil d3no (con / m R
• ' - - . fi? I S H i r A s n n n ! 6 2 . 1 3 dl/dlSOOO 2 6 . 7 3 tir 5 0 0 0 - 1 0 0 0 0
11.14 tren 10000
I
Hien tgi sau 7 ngdy Hinh 14. Dir bSo su thay ddi giS tri theo thdi gian.
Tiip theo, md hinh se tu cap nhat vao cdc gid tri mdi dy bdo dugc d tten thdng qua cdc mit nhap (mit khdng cd cha), rdi thyc hifn vifc suy diin dl nhan kit qua du bao mdi v6 miic dg nhiem ray sau 7 ngay nhu Hinh 14.
HStsafSrnooJidBngftnnBi
^ 62.13 dudsoob a£.n tirsKO-iooao 11.14 Mn 10000
Hlnh 15. Ket qua du b£o d tucmg lai, sau 7 ng&y.
Nhgn xet: Trong md hinh nay, chiing ta thay cac nut nhap da dugc thay bdng cac ket qua dy bdo mdi do xich Markov tta ve. Tir dd, kit qua dy bao sau 7 ngdy Id: nh? chilm 35.08%, ttung binh chilm 52.15%, nhilm ndng chilm 12.77%. So sdnh vdi kit qua miic d0 nhilm riy d thdi diem hifn tai cua md hinh: nh? 8.77%, trung binh 78.39%, nhilm ndng 12.83%, chiing ta thay kit qud da biin dgng theo thdi gian.
V. KET LUAN
Md hinh dy bdo theo thdi gian dugc thiit kl dya tten kien thiic chuyen gia ttong linh bao vf thyc vat va hai cdng cy la xich Markov va mang Bayes. Dilm ndi bat nhat cua md hinh nay la cd the dy bao dugc miic do nhilm ray d nhieu thdi dilm khac nhau d tuong lai ttong qua trinh canh tac liia cua mdi vu miia. Hifn nay, md hinh ndy dang dugc su dung dl thir nghifm dy bdo cho cac vin dl vl dich hai riy nau tten Ilia cua 22 tinh thanh phia Nam la: Dy bdo miic do nhiem ray, dy bdo miic do chdy riy va dy bdo kha nang lan truyen ray d cdc viing lan can.
Trong tuong lai, md hinh ndy cd thl dugc van dung de giai quyet cdc bai toan dy bdo md cac ylu td dau vao tuong tdc vdi nhau theo luat nhan qua, cdc ttang thai cua chiing thda man dieu kifn dung ciia xich Markov yd suy dien dudi dilu kifn ngiu nhien khdng chac chan.
LOI CAM ON
Trudc het, chiing tdi xin cam on chii nhifm de tdi ttgng diem cap Nha nude: "Nghien ciiu xay dyng cdc hf thdng thdng tin hd ttg vifc phdng chdng dich bfnh cay ttdng va thiiy san cho vilng kinh tl ttgng dilm", nia sd: KC.Ol.15/06-10 da t£io dilu kifn thu§n lgi vd hd trg kinh phi cho vifc khao sdt thyc tl dl phuc yu cho vifc nghien ciiu.
Ddng thdi, chiing tdi cung xin giii ldi cam on din GS. Philippe Leray, trudng Dai hgc Bdch khoa Nantes ve sy hd ttg cac ngudn tai lifu chuyen mdn hiiu ich.
Dac bift, chiing tdi chan thdnh cam on cdc chuyen gia d Trung tdm Bao vf Thyc vat Phia Nam: TS. Hd Van Chilli, ThS. Le Qudc Cudng, ThS. Dd Vdn Vin, KS. Nguyen Minh Thu, KS. Nguyen Thj Trang ciing vdi cdc nhan vien d Chi eye Bdo ve Thyc vat Tien Giang va Chi cue Bao yf Thyc vat Ddng Thdp da tich eye hd ttg chiing tdi ve mat nghifp vy ciing nhu vifc thu thap sd lifu phyc vy cho cdng tdc nghien ciiu vd ttien khai cac md hinh du bdo.
TAI LIEU THAM KHAO
[1] Amstnip S. C , B. G. Marcot, and D. C. Douglas. (2007).
Forecasting the range-wide status of polar bears at selected times in the 21st century. Administrative Report: U.S Department of the Interior, U.S. Geological Survey.
[2] Bo Ndng nghi?p va PTNN (2008), S6 lay huong dan phong trie rdy nau truyen bfrih vang liin, liin xoan Id hgi liia, Trung tam Khuyen ndng Qudc gia.
[3] Cecil Huang (1994), "Inference in Belief networlcs: A Procedural Guide", International Joumal of Approximate Reasoning, 11:1-158 Elsevier Science Inc, New York.
[4] Cyc bao v? thyc v?it (2009), Quy^t dinh sS 325/2009/QD- BVTV cong nhgn "Gidi phdp gieo sg dong logt ni ray tren difn rqng de phong bfnh vdng liin, liin xoan Id (VL, LXL) a Ddng bdng song Ciru Long " Id tiin bp khoa hgc ky Ihugt, Bp Ndng nghidp \k Ph^t tri£n ndng thdn.
[5] David Barber (2009), Bayesian Reasoning and Machine Learning, http://www.cs.ucI.ac.Uk/staff/D.Baiber/, UK.
TuyIn tap Cong trinh Nghien ciiu Cdng nghe Thdng tin va Truyin thong 2010
[6] Donald Gross, Carl M. Harris, Wiley (2008), Fundamentals of Queueing Theory. John Wiley & Sons, Inc. New York, NY, USA.
[7] Hd Van Chien, Le Qudc Cudng, Do Van Van, Tdi lifu ve chong ray ciia Trung tam Bdo v$ Thue vdt Phia Nam, Cue Bao vc Thue vat. Bd NN&PTNT, unpublished.
[8] Jeremy Cain (2001), Planning improvements in natural resources management. CEH Wallingford, UK.
[9] Kevin B. Korb, Ann E. Nicholson (2004), Bayesian Artificial Intelligence, Chapman & Hall/CRC Press, UK.
[10] Lukas Sklenar (2004), "Bayesian Belief Network Propagation Engine In Java", University of Kent.
[11] McNay R. S., Marcot B. G., Brumovsky V., and Ellis R.
(2006). "A Bayesian approach to evaluating habitat suitability for woodland caribou in North-central British Columbia", Canadian Joumal of Forest Research 36:3117- 3133, doi: 10.n39/X06-258, NRC Canada.
[12] Nir Friedman, Kevin Murphy, Stuart Russell (1998), Learning the Structure of Dynamic Probabilistic Networks, Computer Science Division, University of California, Berkeley, CA 94720.
[13] Nyberg J. B., Marcot B. G., and Sulyma R. (2006), "Using Bayesian belief networks in adaptive management", Canadian Joumal of Forest Research 36:3104-3116, doi:
10.1139/X06-108, NRC Canada.
[14] Reissig W.H., Heinrichs E.A., Litsinger J.A., Moody K., Fiedler L., Mew W., Barrion A.T. (1993), Huang ddn bi?n phdp tung harp phong trir dich hgi tren lua a Chau A nhiet ddi, NXB Ndng Nghiep.
[15] Robert K. McCann, Bmce G. Marcot, and Rick Ellis, (2006),
"Bayesian belief network: application in ecology and national resource management", CEH - Center for Ecology
& Hydrology, Natural Environment Reseach Council, Can.
J. For. Res. 36: 3053-3062. doi: 10.1139/X06-238, NRC Canada.
[16] Uffe B. Kjaemlff, Anders L. Madsen (2005), Probabilistic Networks - Introduction to Bayesian Networks and Influence Diagrams, Aalborg.
