Optimal Decision-making Criterion Analysis Based On Genetic Algorithms And Its Application In Medicine And Pharmacy

There are a lot of decision-making criterion analysis in the research of medicine and pharmacy, for example, remission and consolidating time for initial treatment of patients with acute leukemia, the optimum cut-off point of diagnostic test, the optimum allocation of public health resources, the optimum condition of animal disease model and the optimal extraction of medicine effective component. Those all belong to the policy-making criterion analysis.Decision-making criterion analysis is to evaluate variables via objective function, and to study the optimal possibilities of different make-ups for decision-making solution set, then to make decision-make according to proper conditions.There is an optimal solution for each single objective problem, but the solution is not of uniqueness for single objective problem with constraints. Direct approach, contour method, the rapid dropping law, enumerating law and so on, there are some local or human-like solution in the solutions.Multi-objective method is to find a set of selectable and non-controllable optimal solution, it is so called 'Pareto Set', by corresponding alternative operations on each sub-objectives. And, Given a set of alternative allocations and a set of individuals, a movement from one allocation to another that can make at least one individual better off, without making any other individual worse off.Decision-makers want to hold more than one solution in practical applications, and in traditional ways, multi-objectives were converted into a single objective or a series of objectives. Such as, Goal Attainment Method, Multiplication and Division Method, the Linear Weighting and the effect method of relate, and these methods try to get a optimal solution for a certain sub-objective, not all sub-objectives, and it can provide the only solution which is always the problem for Operation Researchers.A genetic algorithm (or GA) is a search technique used in computing to find true or approximate solutions to optimization and search problems, which inspired by Darwin's evolution theory. Genetic algorithms are a particular class of evolutionary algorithms that use techniques inspired by evolutionary biology such as inheritance, mutation, selection, and crossover.Genetic algorithms are categorized as global search heuristics, rather than local search, and to provide global solution for the single objective problem. For Multi-objective problems, the populations of solution dataset are evolved and Non-dominated results are found in a Inherent & parallel way. The primary reason for this is their ability to find multiple Pareto-optimal solutions in one single run. Since the principal reason why a problem has a multi-objective formulation is because it is not possible to have a single solution which simultaneously optimizes all objectives, an algorithm that gives a large number of alternative solutions lying on or near the Pareto-optimal front is of great practical value.Genetic Algorithms' theory was introduced in this research, and then a simulation tool of SGALAB was applied into the optimal anemia decision-making criterion analysis and application in Medicine and Pharmacy. SGALAB is a GAs toolbox for MATLAB by Yi Chen who is now doing his research in University of Glasgow, UK. There are 5 parts of this dissertation:Part 1: Single objective GAs simulation and application. The standard test functions are Max f₁(x) = x²,Max f₂(x) = (x-2)², x∈[0,2] and objective results approach or equal to 4, which show that with the function, unique solution has fairly good degree of approximation. That point out that the single objective optimized result has reached level of testing function theory value, and the used SGALAB beta5 procedure is feasible. Since GAs is randomness, that is processed for another times in application, choosing the maximum of the objective function value or the mean as the optimum.On that basis, This chapter focuses study of remission and consolidating time for initial treatment of patients with acute leukemia, who are complete remission and the longest survival time of not less than 24 months. Also, it illustrates optimal extraction of Rutin in more complexity model with variable unconstrained and design-making controlled. The results showed that : Patients can live for up to 78.3 months when remission time is 72.4 months and consolidating is 41.1 months. Initial complete remission of acute leukemia patients could live for no less than 24 months, when remission time is for no less than 19.1 months and consolidation time is for no less than 30.1 months.The study of optimal extraction of Rutin with supercritical CO₂ suggest that : the best extraction condition is: extraction pressure 18Mpa, extraction temperature 45℃modifier used 36ml, Modifier volume fraction of 0.95, static equilibrium time of 30 min, 45 min dynamic extraction time. And Rutin extraction volume can be achieved 5.5023ug/g ,which increased 10.68% than maximum extraction the of 8th test.Part2: Multi-objective GAs effect analysis and procedure available study. The test function f₁(x) = x² ,Max f₂(x) = (x-2)² , x∈[0,2] carrying out two-objective maximum optimization. Result demonstrate Pareto non-dominant results of decision-making variable X are all scattered near the crossing of two functions, the points are optimum selected range for the X ; MOGA NPGA NSGA NSGA II 95%CI contained the intersection point solution, The result of VEGA prefer the maximum. The fore-end of the non-dominant result scatters along the smooth curve; The sub-objective function value and the mean level of Pareto non-dominant solution of the decision-making variable x are separately significant in statistics (F= 0.184 , P< 0.001 ).The solutions searched by NPGA are various and diversity, but MOGA and NSGA accuracy are high, the non-dominant solutions of VEGA beginning and end are more. Point out that: two conflicted objective optimization, first consider MOGA , NPGA , NSGA , secondary chooses NSGA II, and the VEGA result may be a reference. In application, according to the actual, problem, select suitable and satisfied or the mean level as the optimum from the provided Pareto set. Since the searched results accords with testing function mathematics theory, it is considered that the multi-objective GAs procedure is available.Part 3. Application of Multi-objective GAs in optimal diagnostic decision-making criterion. The optimum achieves on the background of maximal sensitivity and specificity in diagnostic test. In traditional ways, the function intersection of sensitivity and specificity will be selected, but in fact, it is some local. The research on MCV optimal decision-making criterion shows there are no significant between different function values and MCV levels, which both including the intersections. But the mutations are different, MOGA and NPGA provide multiplicity solutions. Therefore, if optimize both, MOGA or NPGA is still the fist choice. If the treatment of iron-lacked anemia is effective, less harmful to normal and cost is available, we can select the MOGA 8th plan, in which the specificity is good. And according it , the man whose MCV< 84 fl should be anemia, then its sensitivity is 76.5%, specificity is 50.0%.The decision-making value can also be the mean level calculated by MOGA. Meanwhile, its sensitivity is 61.8%, specificity is 68.2%.Though there are many objective functions, which are no significance, it can be know that grasping main objective can reach the purpose.Diagnostic test decision-making cut-off value is closely related to the costs, and needs to set ROC as the optimal starting point to minimize the cost, and then calculate the decision-making cut-off value through simulation. Because of the great difficulty in doing so, the problem of costs is often neglected in application. My research here has transformed it intoâ‘ the optimal working point of ROC curve line tangent intercept intercept =se- slope·(1 - sp) is at its maximum;â‘¡ROC curve line optimal working point tangent slope is equal to ROC curve line, i.e. constraint = -abs(df(1 - sp) - slope) is the maximum, to get the best match of the two via GAs. The result indicates: when the cost ratio is 0.25, MCV<88.5 fl is diagnosed as anemia, its sensitivity is 94.1 % .specificity is 30.3 %; when the cost ratio0.5, MCV < 82.0 fl diagnosed as anemia, its sensitivity is 70.6%, specificity is 63.6% ; when the costs rate is 0.75, MCV< 78.0 fl is diagnosed as anemia, its sensitivity is 50.0%,specificity is 77.3 %; when the cost ratio in creases, the chosen MCV decision-making level will be low, the sensitivity of the method drops, and there is no much practical value.There are two other methods—finding the derivative of function and bio-normal model. The result of first are similar to GAs and show no statistic significance in most parts, of which MOGA,NPGA,NSGA II are the most satisfactory, the next is NSGA,. To optimize the two conflicted objectives, VEGA is complementary. Since GAs demands no detailed mathematic analysis of the nature the optimal problem, the effect is better than binormal model.Part 4 Application of optimal decision-making criterion about disease prognosis estimation based on multi-objective GAs. It is the study in serous head-hurt mans show that, without considered cost Rate, result of tow-objective is acceptable, and the decision-making cut-off points of CK-BB searched by the five methods are no significance. And the results point that if the damage early time has the better preventive measure, the MOGA 15th plan can be chose. Because when CK-BB> 91U/L, its sensitivity is 85.4%, specificity is 42.1%.When considering the cost Rate, the results searched by MOGA,NPGA,NSGA II are acceptable. The research shows, when cost ratio equal 1.0 and CK-BB>86U/L, its sensitivity is 87.8% and Specificity is 42.1%;when the cost ratio equal 2.0 and CK-BB>233U/L, its sensitivity is 58.5% and Specificity is 84.2%.These conclusions can be referenced by the doctors to evaluate the affection of curing and recovering.Because the incident rate of severe head-damaged patients and observed anemia patients are separately 68.3% and 34.0%, the correspond cost ratio ranges from 0.1 to 4.0 and from 0.1 to 1.5.then known that the value of cost ratio range from 0.5 to 2.5 is available, but we should consider the priority probability in practice to selected a reasonable cost ratio.Part 5 Optimize the extraction medicine effective component. This chapter studies the data of Fructus Schisandrae Chinensis themicrowave-extraction, which including four mutually competed multi-objectives—JinGao rate(%), Schizandrin armor content(%), total Schisandrae fat element content(%),Schizandrin transferred rate(%),and the results show VEGA produces more solutions at the beginning-terminal and gives the max. Then for the problem like it we can choose VEGA first, next NSGA, NSGA II the end. These three methods have always achieved above single objective functional 77% values both in Schizandrin armor content(%) and total Schisandrae fat element content(%),and optimized the condition without making others worse off. We can also get better effect, if the first three evaluations be used as objective function, because the Schizandrin transferred rate (%) could be calculated via Schisandrae. The results including three objectives of NSGA II 11th plan, of which extractions are 50 grams Schisandrae crushed 79 item, 4.5 times of 89% ethyl alcohol joined and 8 minutes under 703W microwave, show that the JinGao rate is 22.06%,Schizandrin armor content is 5.06% and total Schisandrae fat element content is 11.83%.