计算机工程
計算機工程
계산궤공정
COMPUTER ENGINEERING
2014年
7期
198-201
,共4页
方丽英%陈培煜%王普%李爽%杨建栋%万敏
方麗英%陳培煜%王普%李爽%楊建棟%萬敏
방려영%진배욱%왕보%리상%양건동%만민
肿瘤进展%粒子群优化算法%支持向量机%参数寻优%分类模型
腫瘤進展%粒子群優化算法%支持嚮量機%參數尋優%分類模型
종류진전%입자군우화산법%지지향량궤%삼수심우%분류모형
tumor progression%Particle Swarm Optimization(PSO) algorithm%Support Vector Machine(SVM)%parameter optimization%classification model
为研究患者肿瘤进展情况与各项指标之间的关系,以支持向量机(SVM)作为分类模型,根据各项检查指标预测肿瘤进展情况。设计三层粒子群优化算法(tlPSO)对SVM模型进行参数寻优,使用训练集建立分类模型,利用测试集评估模型性能,得到tlPSO-SVM模型。tlPSO算法能有效降低陷入局部最优解的机率,获取全局最优参数,从而使模型具有最优的分类性能。将血常规、中医症候、FACT评分等指标作为输入,肿瘤进展情况作为分类输出,建立分类模型并进行预测。实验结果表明,tlPSO-SVM模型准确率较高,具有较好的分类性能。
為研究患者腫瘤進展情況與各項指標之間的關繫,以支持嚮量機(SVM)作為分類模型,根據各項檢查指標預測腫瘤進展情況。設計三層粒子群優化算法(tlPSO)對SVM模型進行參數尋優,使用訓練集建立分類模型,利用測試集評估模型性能,得到tlPSO-SVM模型。tlPSO算法能有效降低陷入跼部最優解的機率,穫取全跼最優參數,從而使模型具有最優的分類性能。將血常規、中醫癥候、FACT評分等指標作為輸入,腫瘤進展情況作為分類輸齣,建立分類模型併進行預測。實驗結果錶明,tlPSO-SVM模型準確率較高,具有較好的分類性能。
위연구환자종류진전정황여각항지표지간적관계,이지지향량궤(SVM)작위분류모형,근거각항검사지표예측종류진전정황。설계삼층입자군우화산법(tlPSO)대SVM모형진행삼수심우,사용훈련집건립분류모형,이용측시집평고모형성능,득도tlPSO-SVM모형。tlPSO산법능유효강저함입국부최우해적궤솔,획취전국최우삼수,종이사모형구유최우적분류성능。장혈상규、중의증후、FACT평분등지표작위수입,종류진전정황작위분류수출,건립분류모형병진행예측。실험결과표명,tlPSO-SVM모형준학솔교고,구유교호적분류성능。
For researching the relation between the lung cancers’ tumor progression and the medical factors, taking Support Vector Machine(SVM) as the classification model. Firstly, utilizing the improved Particle Swarm Optimization(PSO) algorithm-three layers Particle Swarm Optimization(tlPSO), to optimize the SVM parameters; Secondly, establishing the classification model; Finally, using the test set to evaluate the model. The tlPSO algorithm reduces the probability of falling into local optimal solution by the algorithm, obtained the global optimal conclusion and makes the model with optimal performance. Choosing blood routine, the TCM symptoms and the FACT value as the input of the experiment, the tumor progression as the classification output, the classification model is established for prognosis. From the experiment, the improved PSO is superior in finding the optimal parameters and improved the accuracy of classification model, and the proposed tlPSO-SVM model has better classification performance.