File size: 5,283 Bytes
dfcdf7b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import os
import sys
import matplotlib.pyplot as plt
import numpy as np
import random
import jprops
from random import randint
from matumizi.util import *
from matumizi.mlutil import *

"""
Markov chain classifier
"""
class MarkovChainClassifier():
	def __init__(self, configFile):
		"""
		constructor
		
		Parameters
			configFile: config file path
		"""
		defValues = {}
		defValues["common.model.directory"] = ("model", None)
		defValues["common.model.file"] = (None, None)
		defValues["common.verbose"] = (False, None)
		defValues["common.states"] = (None, "missing state list")
		defValues["train.data.file"] = (None, "missing training data file")
		defValues["train.data.class.labels"] = (["F", "T"], None)
		defValues["train.data.key.len"] = (1, None)
		defValues["train.model.save"] = (False, None)
		defValues["train.score.method"] = ("accuracy", None)
		defValues["predict.data.file"] = (None, None)
		defValues["predict.use.saved.model"] = (True, None)
		defValues["predict.log.odds.threshold"] = (0, None)
		defValues["validate.data.file"] = (None, "missing validation data file")
		defValues["validate.use.saved.model"] = (False, None)
		defValues["valid.accuracy.metric"] = ("acc", None)
		self.config = Configuration(configFile, defValues)
		
		self.stTranPr = dict()
		self.clabels = self.config.getStringListConfig("train.data.class.labels")[0]
		self.states = self.config.getStringListConfig("common.states")[0]
		self.nstates = len(self.states)
		for cl in self.clabels:
			stp = np.ones((self.nstates,self.nstates))
			self.stTranPr[cl] = stp
		
	def train(self):
		"""
		train model
		"""	
		#state transition matrix
		tdfPath = self.config.getStringConfig("train.data.file")[0]
		klen = self.config.getIntConfig("train.data.key.len")[0]
		for rec in fileRecGen(tdfPath):
			cl = rec[klen]
			rlen = len(rec)
			for i in range(klen+1, rlen-1, 1):
				fst = self.states.index(rec[i])
				tst = self.states.index(rec[i+1])
				self.stTranPr[cl][fst][tst] += 1
		
		#normalize to probability
		for cl in self.clabels:
			stp = self.stTranPr[cl]
			for i in range(self.nstates):
				s = stp[i].sum()
				r = stp[i] / s
				stp[i] = r
		
		#save		
		if 	self.config.getBooleanConfig("train.model.save")[0]:
			mdPath = self.config.getStringConfig("common.model.directory")[0]
			assert os.path.exists(mdPath), "model save directory does not exist"
			mfPath = self.config.getStringConfig("common.model.file")[0]
			mfPath = os.path.join(mdPath, mfPath)

			with open(mfPath, "w") as fh:
				for cl in self.clabels:
					fh.write("label:" + cl +"\n")
					stp = self.stTranPr[cl]
					for r in stp:
						rs = ",".join(toStrList(r, 6)) + "\n"
						fh.write(rs)

	def validate(self):
		"""
		validate using  model
		"""	
		useSavedModel = self.config.getBooleanConfig("predict.use.saved.model")[0]
		if useSavedModel:
			self.__restoreModel()
		else:
			self.train() 
			
		vdfPath = self.config.getStringConfig("validate.data.file")[0]	
		accMetric = self.config.getStringConfig("valid.accuracy.metric")[0]
		
		yac, ypr = self.__getPrediction(vdfPath, True)
		if type(self.clabels[0]) == str:
			yac = self.__toIntClabel(yac)
			ypr = self.__toIntClabel(ypr)
		score = perfMetric(accMetric, yac, ypr)
		print(formatFloat(3, score, "perf score"))

			
	def predict(self):
		"""
		predict using  model
		"""	
		useSavedModel = self.config.getBooleanConfig("predict.use.saved.model")[0]
		if useSavedModel:
			self.__restoreModel()
		else:
			self.train() 
			
		#predict
		pdfPath = self.config.getStringConfig("predict.data.file")[0]
		_ , ypr = self.__getPrediction(pdfPath)
		return ypr
		
	def __restoreModel(self):
		"""
		restore model
		"""
		mdPath = self.config.getStringConfig("common.model.directory")[0]
		assert os.path.exists(mdPath), "model save directory does not exist"
		mfPath = self.config.getStringConfig("common.model.file")[0]
		mfPath = os.path.join(mdPath, mfPath)
		stp = None
		cl = None
		for rec in fileRecGen(mfPath):
			if len(rec) == 1:
				if stp is not None:
					stp = np.array(stp)
					self.stTranPr[cl] = stp
				cl = rec[0].split(":")[1]
				stp = list()
			else:
				frec = asFloatList(rec)
				stp.append(frec)
				
		stp = np.array(stp)
		self.stTranPr[cl] = stp
				
	def __getPrediction(self, fpath, validate=False):
		"""
		get predictions
		
		Parameters
			fpath : data file path
			validate: True if validation
		"""
	
		nc = self.clabels[0]
		pc = self.clabels[1]
		thold = self.config.getFloatConfig("predict.log.odds.threshold")[0]
		klen = self.config.getIntConfig("train.data.key.len")[0]
		offset = klen+1 if validate else klen
		ypr = list()
		yac = list()
		for rec in fileRecGen(fpath):
			lodds = 0
			rlen = len(rec)
			for i in range(offset, rlen-1, 1):
				fst = self.states.index(rec[i])
				tst = self.states.index(rec[i+1])
				odds = self.stTranPr[pc][fst][tst] / self.stTranPr[nc][fst][tst]
				lodds += math.log(odds)
			prc = pc if lodds > thold else nc
			ypr.append(prc)
			if validate:
				yac.append(rec[klen])
			else:
				recp = prc + "\t" + ",".join(rec)
				print(recp)

		re = (yac, ypr)
		return re
	
	def __toIntClabel(self, labels):
		"""
		convert string class label to int
		
		Parameters
			labels : class label values
		"""
		return list(map(lambda l : self.clabels.index(l), labels))