├── markov.py
├── readme.md
└── trajectory100000.csv


/markov.py:
--------------------------------------------------------------------------------
  1 | import csv
  2 | import operator
  3 | 
  4 | 
  5 | class Markov:
  6 |     markov_matrix = {}
  7 |     his_count = 1
  8 |     pre_count = 3
  9 |     test_count = 3
 10 |     tol = 3
 11 | 
 12 |     def __init__(self, his_count=1, pre_count=3, test_count=3, tol=3):
 13 |         """
 14 |         :param his_count: history point(s) count.
 15 |         :param pre_count: prediction point(s) count.
 16 |         :param test_count: point(s) count remain for test.
 17 |         :param tol: int, tolerant point(s) count of prediction.
 18 |         """
 19 |         self.his_count = his_count
 20 |         self.pre_count = pre_count
 21 |         self.test_count = test_count
 22 |         self.tol = tol
 23 | 
 24 |     def get_predict(self, his_seq: tuple):
 25 |         """
 26 |         Get prediction from a history sequence.
 27 | 
 28 |         :param his_seq: tuple, history sequence.
 29 |         :return: list, containing count of tolerant point(s) as a prediction set.
 30 |         """
 31 |         result = []
 32 |         predict_list = self.markov_matrix[his_seq]
 33 |         for i in range(self.pre_count):
 34 |             pre_i_dict = predict_list[i]
 35 |             sorted_i_list = sorted(pre_i_dict.items(), key=operator.itemgetter(1))
 36 |             sorted_i_list.reverse()
 37 |             result_i = []
 38 |             for j in range(self.tol):
 39 |                 try:
 40 |                     result_i.append(sorted_i_list[j][0])
 41 |                 except IndexError:
 42 |                     break
 43 |             result.append(result_i)
 44 |         return result
 45 | 
 46 |     def read_csv(self, file_dir: str):
 47 |         """
 48 |         Read sequences from one csv file.
 49 | 
 50 |         :param file_dir: str, file directory of the csv file.
 51 |         """
 52 |         with open(file_dir, 'r') as file:
 53 |             reader = csv.reader(file, delimiter=',')
 54 |             for row in reader:
 55 |                 self.add_row(row)
 56 | 
 57 |     def add_row(self, row: list):
 58 |         """
 59 |         Add one rows to markov matrix.
 60 | 
 61 |         :param row: list, containing one row of training sequence.
 62 |         """
 63 |         input_row = row[:-self.test_count]
 64 |         for i in range(len(input_row) - (self.his_count + self.pre_count) + 1):
 65 |             his_seq = tuple(input_row[i:i + self.his_count])
 66 |             pre_seq = tuple(input_row[i + self.his_count:i + self.his_count + self.pre_count])
 67 |             self.add_seq(his_seq, pre_seq)
 68 | 
 69 |     def add_seq(self, his_seq: tuple, pre_seq: tuple):
 70 |         """
 71 |         Add a new sequence to markov matrix.
 72 | 
 73 |         :param his_seq: tuple, containing one history sequence.
 74 |         :param pre_seq: tuple, containing one prediction sequence.
 75 |         """
 76 |         try:
 77 |             history_row = self.markov_matrix[his_seq]
 78 | 
 79 |             for i in range(self.pre_count):
 80 |                 try:
 81 |                     predict_i_count = history_row[i][pre_seq[i]]
 82 |                     self.markov_matrix[his_seq][i][pre_seq[i]] = predict_i_count + 1
 83 |                 except KeyError:
 84 |                     self.markov_matrix[his_seq][i][pre_seq[i]] = 1
 85 |         except KeyError:
 86 |             self.markov_matrix[his_seq] = []
 87 |             for point in pre_seq:
 88 |                 self.markov_matrix[his_seq].append({point: 1})
 89 | 
 90 |     def test_seq(self, his_seq: tuple, real_seq: tuple):
 91 |         """
 92 |         Use current markov matrix to run a test and return accuracy.
 93 | 
 94 |         :param his_seq: tuple, history sequence used to predict next N sequence.
 95 |         :param real_seq: tuple, correct sequence used to compare to prediction.
 96 |         :return: list of int, correct count of this sequence, with each point in time series separated.
 97 |         """
 98 |         result = [0] * self.pre_count
 99 |         try:
100 |             predict_sequence = self.get_predict(his_seq)
101 |             for i in range(self.pre_count):
102 |                 if real_seq[i] in predict_sequence[i]:
103 |                     result[i] += 1
104 |         except KeyError:
105 |             pass
106 |         return result
107 | 
108 |     def test_csv(self, file_dir: str):
109 |         """
110 |         Use all test data from a csv file to test markov model.
111 | 
112 |         :param file_dir: str, file directory of the csv file.
113 |         :return: float, accuracy calculated using whole file's data.
114 |         """
115 |         aggregate = [0] * self.pre_count
116 |         row_count = 0
117 |         with open(file_dir, 'r') as file:
118 |             reader = csv.reader(file, delimiter=',')
119 |             for row in reader:
120 |                 row_count += 1
121 |                 aggregate = [sum(x) for x in
122 |                              zip(self.test_seq(tuple(row[-(self.his_count + self.test_count):-self.test_count]),
123 |                                                tuple(row[-self.test_count:])), aggregate)]
124 |         return [x/row_count for x in aggregate]
125 | 
126 |     def write_pre(self, test_dir: str, result_dir: str):
127 |         """
128 |         Predict sequences using test data and write predict result into a csv file.
129 | 
130 |         :param test_dir: str, test data's file directory
131 |         :param result_dir: str, predict result's file directory
132 |         """
133 |         test_file = open(test_dir, 'r')
134 |         result_file = open(result_dir, 'w', newline='')
135 |         reader = csv.reader(test_file, delimiter=',')
136 |         writer = csv.writer(result_file, delimiter=',')
137 |         for row in reader:
138 |             try:
139 |                 predict_sequence = self.get_predict(tuple(row[-(self.his_count + self.test_count):-self.test_count]))
140 |                 max_seq = []
141 |                 for point_list in predict_sequence:
142 |                     if len(point_list) > 0:
143 |                         max_seq.append(point_list[0])
144 |                     else:
145 |                         max_seq.append(None)
146 |                 writer.writerow(max_seq)
147 |             except KeyError:
148 |                 writer.writerow([None] * self.pre_count)
149 |         test_file.close()
150 |         result_file.close()
151 | 
152 | 
153 | if __name__ == '__main__':
154 |     markov = Markov(his_count=1)
155 |     markov.read_csv('trajectory100000.csv')
156 |     print(markov.test_csv('trajectory100000.csv'))
157 |     markov.write_pre('trajectory100000.csv', 'result.csv')
158 |     print()
159 | 


--------------------------------------------------------------------------------
/readme.md:
--------------------------------------------------------------------------------
 1 | # 多候选马尔可夫轨迹预测
 2 | 
 3 | ## 功能
 4 | 
 5 | - 导入csv数据并计算马尔可夫矩阵
 6 | - 根据给定历史序列给出预测
 7 | - 每个点的预测都能给出多个候选值
 8 | - 历史轨迹长度、预测轨迹长度和候选值数量均可调整
 9 | 
10 | ## 使用
11 | 
12 | ### 初始化
13 | 
14 |     markov = Markov(his_count=1, pre_count=3, test_count=3, tol=3)
15 | 
16 | - `his_count`为历史轨迹长度
17 | - `pre_count`为预测轨迹长度
18 | - `test_count`为csv文件中每一行数据需要留作测试集的长度（取最后`test_count`数量的点作测试集）
19 | - `tol`为候选值数量
20 | 
21 | ### 读入csv文件
22 | 
23 |     markov.read_csv('trajectory100000.csv')
24 | 
25 | ### 使用csv文件数据进行测试
26 | 
27 |     acc = markov.test_csv('trajectory100000.csv')
28 | 
29 | ### 导出预测序列
30 | 
31 |     markov.write_pre('trajectory100000.csv', 'result.csv')
32 | 


--------------------------------------------------------------------------------