import pandas as pd
import numpy as np
class NaiveBayes:
def __init__(self, data, lam=0):
df = pd.DataFrame(data)
dim = df.shape[1]
self.y_p = df[dim-1].groupby(df[dim-1]).count()+lam
self.y_p /= self.y_p.sum()
self.cb = []
for i in range(dim-1):
xi_p = pd.crosstab(df[i], df[dim-1])+lam
self.cb.append(xi_p/xi_p.sum())
def predict(self, x):
labels = self.y_p.index
res = pd.Series(np.zeros(labels.size), index=labels)
for y in labels:
res[y] = self.y_p[y]
for i in range(len(x)):
try:
res[y] *= self.cb[i][y][x[i]]
except KeyError as e:
print('KeyError!', e)
return None
print(res)
return res.idxmax()
if __name__ == "__main__":
df = pd.read_csv('../data/4-1.txt', header=None)
NB = NaiveBayes(df, 1)
res = NB.predict([2, 'S'])
print('res:', res)
