A Hadoop-Based Method to Predict Potential Effective Drug Combination

Algorithm 1

The workflow of the scalable version of the Naïve Bayesian algorithm implemented by MapReduce.

Step  1.

map:

 foreach training sample: (𝐶_𝑖,𝑋_𝑗)

  𝑒𝑚𝑖𝑡𝐶_𝑖,1

reduce:

  𝑒𝑚𝑖𝑡𝐶_𝑖,sum𝐶_𝑖

Step  2.

map:

 foreach training sample: (𝐶_𝑖,𝑋_𝑗)

  foreach feature 𝑓_𝑘

   𝑒𝑚𝑖𝑡𝑓_𝑘∣𝐶_𝑖,1

reduce:

 foreach class

  𝑓𝑒𝑚𝑖𝑡^𝑘∣𝐶_𝑖sum(𝐶_𝑖),sum𝑓_𝑘∣𝐶_𝑖

Step  3.

map:

 foreach class 𝐶_𝑖

  foreach testing sample: (𝑋_𝑗)

   𝑒𝑚𝑖𝑡𝑋_𝑗,𝑃𝐶_𝑖∏𝑃𝑋_𝑗∣𝐶_𝑖

reduce:

 𝑒𝑚𝑖𝑡𝑋_𝑗,argmax𝑃𝐶_𝑖∏𝑃𝑋_𝑗∣𝐶_𝑖