12、微阵列癌症数据的基因标记物筛选与分类模型研究

微阵列癌症数据的基因标记物筛选与分类模型研究

1. 数据构建与特征选择规则

真实数据集由 6 个样本中 50 个正常基因值开始，随后是 50 个癌症基因值构成。为了筛选出有价值的基因标记物并计算错误率，制定了一系列 KP 规则，具体如下表所示：
| No. | Feature selection (FS): KP rules to get marker genes and calculate error rates (priority based on the rules No.) | KP rules name |
| — | — | — |
| m1 | s →find(M(:) == i), table(i) →lenght(s), {1 ≤i ≤100} | Rewriting |
| m2 | markergenes(i, it + 1) →i{table(i) > it, 1 ≤i ≤100, 1 ≤it ≤n} | Rewriting |
| m3 | y(i, 1) →+1 & maxj →maxj + 1, {1 ≤i ≤50} | Rewriting |
| m4 | y(i, 1) →−1 & maxf →maxf + 1, {51 ≤i ≤100} | Rewriting |
| m5 | BadGenes(it + 1, 1) →maxf | Rewriting |
| m6 | Y (j, 1) →+1, {1 ≤j ≤maxj ∗3} Y (j, 1) →−1, {maxj ∗3 + 1 ≤j ≤ maxj ∗3 + maxf ∗3} | Rewriting |
| m7 | w