通过理想值强度值分配隐马尔可夫模型的状态。
Assigning states of Hidden Markov Models by idealized values intensity values.
我的数据是运行石榴HMM（http://pomegranate.readthedocs.io/en/latest/HiddenMarkovModel.html），我将结果加载到Pandas DF中，并将理想强度定义为该状态下的所有点数：df["hmm_idealized"] = df.groupby(["hmm_state"],as_index = False)["Raw"].transform("median")。示例数据：
+-----+-----------------+-------------+------------+
| | hmm_idealized | hmm_state | hmm_diff |
|-----+-----------------+-------------+------------|
| 0 | 99862 | 3 | nan |
| 1 | 99862 | 3 | 0 |
| 2 | 99862 | 3 | 0 |
| 3 | 99862 | 3 | 0 |
| 4 | 99862 | 3 | 0 |
| 5 | 99862 | 3 | 0 |
| 6 | 117759 | 4 | 1 |
| 7 | 117759 | 4 | 0 |
| 8 | 117759 | 4 | 0 |
| 9 | 117759 | 4 | 0 |
| 10 | 117759 | 4 | 0 |
| 11 | 117759 | 4 | 0 |
| 12 | 117759 | 4 | 0 |
| 13 | 117759 | 4 | 0 |
| 14 | 124934 | 2 | -2 |
| 15 | 124934 | 2 | 0 |
| 16 | 124934 | 2 | 0 |
| 17 | 124934 | 2 | 0 |
| 18 | 124934 | 2 | 0 |
| 19 | 117759 | 4 | 2 |
| 20 | 117759 | 4 | 0 |
| 21 | 117759 | 4 | 0 |
| 22 | 117759 | 4 | 0 |
| 23 | 117759 | 4 | 0 |
| 24 | 117759 | 4 | 0 |
| 25 | 117759 | 4 | 0 |
| 26 | 117759 | 4 | 0 |
| 27 | 117759 | 4 | 0 |
| 28 | 117759 | 4 | 0 |
| 29 | 117759 | 4 | 0 |
| 30 | 117759 | 4 | 0 |
| 31 | 117759 | 4 | 0 |
| 32 | 117759 | 4 | 0 |
| 33 | 117759 | 4 | 0 |
| 34 | 117759 | 4 | 0 |
| 35 | 117759 | 4 | 0 |
| 36 | 117759 | 4 | 0 |
| 37 | 117759 | 4 | 0 |
| 38 | 117759 | 4 | 0 |
| 39 | 117759 | 4 | 0 |
| 40 | 106169 | 1 | -3 |
| 41 | 106169 | 1 | 0 |
| 42 | 106169 | 1 | 0 |
| 43 | 106169 | 1 | 0 |
| 44 | 106169 | 1 | 0 |
| 45 | 106169 | 1 | 0 |
| 46 | 106169 | 1 | 0 |
| 47 | 106169 | 1 | 0 |
| 48 | 106169 | 1 | 0 |
| 49 | 106169 | 1 | 0 |
| 50 | 106169 | 1 | 0 |
| 51 | 106169 | 1 | 0 |
| 52 | 106169 | 1 | 0 |
| 53 | 106169 | 1 | 0 |
| 54 | 106169 | 1 | 0 |
| 55 | 106169 | 1 | 0 |
| 56 | 106169 | 1 | 0 |
| 57 | 106169 | 1 | 0 |
| 58 | 106169 | 1 | 0 |
| 59 | 106169 | 1 | 0 |
| 60 | 106169 | 1 | 0 |
| 61 | 106169 | 1 | 0 |
| 62 | 106169 | 1 | 0 |
| 63 | 106169 | 1 | 0 |
| 64 | 106169 | 1 | 0 |
| 65 | 106169 | 1 | 0 |
| 66 | 106169 | 1 | 0 |
| 67 | 106169 | 1 | 0 |
| 68 | 106169 | 1 | 0 |
| 69 | 106169 | 1 | 0 |
| 70 | 106169 | 1 | 0 |
| 71 | 106169 | 1 | 0 |
| 72 | 106169 | 1 | 0 |
| 73 | 106169 | 1 | 0 |
| 74 | 106169 | 1 | 0 |
| 75 | 99862 | 3 | 2 |
| 76 | 99862 | 3 | 0 |
| 77 | 99862 | 3 | 0 |
| 78 | 99862 | 3 | 0 |
| 79 | 99862 | 3 | 0 |
| 80 | 99862 | 3 | 0 |
| 81 | 99862 | 3 | 0 |
| 82 | 99862 | 3 | 0 |
| 83 | 99862 | 3 | 0 |
| 84 | 99862 | 3 | 0 |
| 85 | 99862 | 3 | 0 |
| 86 | 99862 | 3 | 0 |
| 87 | 99862 | 3 | 0 |
| 88 | 99862 | 3 | 0 |
| 89 | 99862 | 3 | 0 |
| 90 | 99862 | 3 | 0 |
| 91 | 99862 | 3 | 0 |
| 92 | 99862 | 3 | 0 |
| 93 | 99862 | 3 | 0 |
| 94 | 99862 | 3 | 0 |
| 95 | 99862 | 3 | 0 |
| 96 | 99862 | 3 | 0 |
| 97 | 99862 | 3 | 0 |
| 98 | 99862 | 3 | 0 |
| 99 | 99862 | 3 | 0 |
| 100 | 99862 | 3 | 0 |
| 101 | 99862 | 3 | 0 |
| 102 | 99862 | 3 | 0 |
| 103 | 99862 | 3 | 0 |
| 104 | 99862 | 3 | 0 |
| 105 | 99862 | 3 | 0 |
| 106 | 99862 | 3 | 0 |
| 107 | 99862 | 3 | 0 |
| 108 | 94127 | 0 | -3 |
| 109 | 94127 | 0 | 0 |
| 110 | 94127 | 0 | 0 |
| 111 | 94127 | 0 | 0 |
| 112 | 94127 | 0 | 0 |
| 113 | 94127 | 0 | 0 |
| 114 | 94127 | 0 | 0 |
| 115 | 94127 | 0 | 0 |
| 116 | 94127 | 0 | 0 |
| 117 | 94127 | 0 | 0 |
| 118 | 94127 | 0 | 0 |
| 119 | 94127 | 0 | 0 |
| 120 | 94127 | 0 | 0 |
| 121 | 94127 | 0 | 0 |
| 122 | 94127 | 0 | 0 |
| 123 | 94127 | 0 | 0 |
| 124 | 94127 | 0 | 0 |
| 125 | 94127 | 0 | 0 |
| 126 | 94127 | 0 | 0 |
| 127 | 94127 | 0 | 0 |
| 128 | 94127 | 0 | 0 |
| 129 | 94127 | 0 | 0 |
| 130 | 94127 | 0 | 0 |
| 131 | 94127 | 0 | 0 |
| 132 | 94127 | 0 | 0 |
| 133 | 94127 | 0 | 0 |
| 134 | 94127 | 0 | 0 |
| 135 | 94127 | 0 | 0 |
| 136 | 94127 | 0 | 0 |
| 137 | 94127 | 0 | 0 |
| 138 | 94127 | 0 | 0 |
| 139 | 94127 | 0 | 0 |
| 140 | 94127 | 0 | 0 |
| 141 | 94127 | 0 | 0 |
| 142 | 94127 | 0 | 0 |
| 143 | 94127 | 0 | 0 |
| 144 | 94127 | 0 | 0 |
| 145 | 94127 | 0 | 0 |
| 146 | 94127 | 0 | 0 |
| 147 | 94127 | 0 | 0 |
| 148 | 94127 | 0 | 0 |
| 149 | 94127 | 0 | 0 |
| 150 | 94127 | 0 | 0 |
| 151 | 94127 | 0 | 0 |
| 152 | 94127 | 0 | 0 |
| 153 | 94127 | 0 | 0 |
| 154 | 94127 | 0 | 0 |
| 155 | 94127 | 0 | 0 |
| 156 | 94127 | 0 | 0 |
| 157 | 94127 | 0 | 0 |
| 158 | 94127 | 0 | 0 |
| 159 | 94127 | 0 | 0 |
| 160 | 94127 | 0 | 0 |
| 161 | 94127 | 0 | 0 |
| 162 | 94127 | 0 | 0 |
| 163 | 94127 | 0 | 0 |
| 164 | 94127 | 0 | 0 |
| 165 | 94127 | 0 | 0 |
| 166 | 94127 | 0 | 0 |
| 167 | 94127 | 0 | 0 |
| 168 | 94127 | 0 | 0 |
| 169 | 94127 | 0 | 0 |
| 170 | 94127 | 0 | 0 |
| 171 | 94127 | 0 | 0 |
| 172 | 94127 | 0 | 0 |
| 173 | 94127 | 0 | 0 |
| 174 | 94127 | 0 | 0 |
| 175 | 94127 | 0 | 0 |
| 176 | 94127 | 0 | 0 |
| 177 | 94127 | 0 | 0 |
| 178 | 94127 | 0 | 0 |
| 179 | 94127 | 0 | 0 |
| 180 | 94127 | 0 | 0 |
| 181 | 94127 | 0 | 0 |
| 182 | 94127 | 0 | 0 |
| 183 | 94127 | 0 | 0 |
| 184 | 94127 | 0 | 0 |
| 185 | 94127 | 0 | 0 |
| 186 | 94127 | 0 | 0 |
| 187 | 94127 | 0 | 0 |
| 188 | 94127 | 0 | 0 |
| 189 | 94127 | 0 | 0 |
| 190 | 94127 | 0 | 0 |
| 191 | 94127 | 0 | 0 |
| 192 | 94127 | 0 | 0 |
| 193 | 94127 | 0 | 0 |
| 194 | 94127 | 0 | 0 |
| 195 | 94127 | 0 | 0 |
| 196 | 94127 | 0 | 0 |
| 197 | 94127 | 0 | 0 |
| 198 | 94127 | 0 | 0 |
| 199 | 94127 | 0 | 0 |
| 200 | 94127 | 0 | 0 |
+-----+-----------------+-------------+------------+
在分析结果的时候，我想统计模型增加的次数。我想使用 df.where(data.hmm_diff > 0).count() 函数来读取我递增一个状态的次数。但是，增加有时会跨越两个状态（即它会跳过中间状态），所以我需要通过对理想化值进行排序来重新分配 HMM 状态标签，以便最低状态为 0，最高状态为 4，等等。是否有一种根据理想化强度将 hmm_state 标签从任意重新分配到 的方法？
例如，标记为“1”的 hmm_state 位于 hmm_state 3 和 4
之间
看起来你只需要像这样定义一个排序的 HMM 状态：
state_orders = {v: i for i, v in enumerate(sorted(df.hmm_idealized.unique()))}
df['sorted_state'] = df.hmm_idealized.map(state_orders)
然后你可以像你在问题中所做的那样继续，但是在这个专栏上做一个差异，并计算它的跳跃。
我的数据是运行石榴HMM（http://pomegranate.readthedocs.io/en/latest/HiddenMarkovModel.html），我将结果加载到Pandas DF中，并将理想强度定义为该状态下的所有点数：df["hmm_idealized"] = df.groupby(["hmm_state"],as_index = False)["Raw"].transform("median")。示例数据：
+-----+-----------------+-------------+------------+
| | hmm_idealized | hmm_state | hmm_diff |
|-----+-----------------+-------------+------------|
| 0 | 99862 | 3 | nan |
| 1 | 99862 | 3 | 0 |
| 2 | 99862 | 3 | 0 |
| 3 | 99862 | 3 | 0 |
| 4 | 99862 | 3 | 0 |
| 5 | 99862 | 3 | 0 |
| 6 | 117759 | 4 | 1 |
| 7 | 117759 | 4 | 0 |
| 8 | 117759 | 4 | 0 |
| 9 | 117759 | 4 | 0 |
| 10 | 117759 | 4 | 0 |
| 11 | 117759 | 4 | 0 |
| 12 | 117759 | 4 | 0 |
| 13 | 117759 | 4 | 0 |
| 14 | 124934 | 2 | -2 |
| 15 | 124934 | 2 | 0 |
| 16 | 124934 | 2 | 0 |
| 17 | 124934 | 2 | 0 |
| 18 | 124934 | 2 | 0 |
| 19 | 117759 | 4 | 2 |
| 20 | 117759 | 4 | 0 |
| 21 | 117759 | 4 | 0 |
| 22 | 117759 | 4 | 0 |
| 23 | 117759 | 4 | 0 |
| 24 | 117759 | 4 | 0 |
| 25 | 117759 | 4 | 0 |
| 26 | 117759 | 4 | 0 |
| 27 | 117759 | 4 | 0 |
| 28 | 117759 | 4 | 0 |
| 29 | 117759 | 4 | 0 |
| 30 | 117759 | 4 | 0 |
| 31 | 117759 | 4 | 0 |
| 32 | 117759 | 4 | 0 |
| 33 | 117759 | 4 | 0 |
| 34 | 117759 | 4 | 0 |
| 35 | 117759 | 4 | 0 |
| 36 | 117759 | 4 | 0 |
| 37 | 117759 | 4 | 0 |
| 38 | 117759 | 4 | 0 |
| 39 | 117759 | 4 | 0 |
| 40 | 106169 | 1 | -3 |
| 41 | 106169 | 1 | 0 |
| 42 | 106169 | 1 | 0 |
| 43 | 106169 | 1 | 0 |
| 44 | 106169 | 1 | 0 |
| 45 | 106169 | 1 | 0 |
| 46 | 106169 | 1 | 0 |
| 47 | 106169 | 1 | 0 |
| 48 | 106169 | 1 | 0 |
| 49 | 106169 | 1 | 0 |
| 50 | 106169 | 1 | 0 |
| 51 | 106169 | 1 | 0 |
| 52 | 106169 | 1 | 0 |
| 53 | 106169 | 1 | 0 |
| 54 | 106169 | 1 | 0 |
| 55 | 106169 | 1 | 0 |
| 56 | 106169 | 1 | 0 |
| 57 | 106169 | 1 | 0 |
| 58 | 106169 | 1 | 0 |
| 59 | 106169 | 1 | 0 |
| 60 | 106169 | 1 | 0 |
| 61 | 106169 | 1 | 0 |
| 62 | 106169 | 1 | 0 |
| 63 | 106169 | 1 | 0 |
| 64 | 106169 | 1 | 0 |
| 65 | 106169 | 1 | 0 |
| 66 | 106169 | 1 | 0 |
| 67 | 106169 | 1 | 0 |
| 68 | 106169 | 1 | 0 |
| 69 | 106169 | 1 | 0 |
| 70 | 106169 | 1 | 0 |
| 71 | 106169 | 1 | 0 |
| 72 | 106169 | 1 | 0 |
| 73 | 106169 | 1 | 0 |
| 74 | 106169 | 1 | 0 |
| 75 | 99862 | 3 | 2 |
| 76 | 99862 | 3 | 0 |
| 77 | 99862 | 3 | 0 |
| 78 | 99862 | 3 | 0 |
| 79 | 99862 | 3 | 0 |
| 80 | 99862 | 3 | 0 |
| 81 | 99862 | 3 | 0 |
| 82 | 99862 | 3 | 0 |
| 83 | 99862 | 3 | 0 |
| 84 | 99862 | 3 | 0 |
| 85 | 99862 | 3 | 0 |
| 86 | 99862 | 3 | 0 |
| 87 | 99862 | 3 | 0 |
| 88 | 99862 | 3 | 0 |
| 89 | 99862 | 3 | 0 |
| 90 | 99862 | 3 | 0 |
| 91 | 99862 | 3 | 0 |
| 92 | 99862 | 3 | 0 |
| 93 | 99862 | 3 | 0 |
| 94 | 99862 | 3 | 0 |
| 95 | 99862 | 3 | 0 |
| 96 | 99862 | 3 | 0 |
| 97 | 99862 | 3 | 0 |
| 98 | 99862 | 3 | 0 |
| 99 | 99862 | 3 | 0 |
| 100 | 99862 | 3 | 0 |
| 101 | 99862 | 3 | 0 |
| 102 | 99862 | 3 | 0 |
| 103 | 99862 | 3 | 0 |
| 104 | 99862 | 3 | 0 |
| 105 | 99862 | 3 | 0 |
| 106 | 99862 | 3 | 0 |
| 107 | 99862 | 3 | 0 |
| 108 | 94127 | 0 | -3 |
| 109 | 94127 | 0 | 0 |
| 110 | 94127 | 0 | 0 |
| 111 | 94127 | 0 | 0 |
| 112 | 94127 | 0 | 0 |
| 113 | 94127 | 0 | 0 |
| 114 | 94127 | 0 | 0 |
| 115 | 94127 | 0 | 0 |
| 116 | 94127 | 0 | 0 |
| 117 | 94127 | 0 | 0 |
| 118 | 94127 | 0 | 0 |
| 119 | 94127 | 0 | 0 |
| 120 | 94127 | 0 | 0 |
| 121 | 94127 | 0 | 0 |
| 122 | 94127 | 0 | 0 |
| 123 | 94127 | 0 | 0 |
| 124 | 94127 | 0 | 0 |
| 125 | 94127 | 0 | 0 |
| 126 | 94127 | 0 | 0 |
| 127 | 94127 | 0 | 0 |
| 128 | 94127 | 0 | 0 |
| 129 | 94127 | 0 | 0 |
| 130 | 94127 | 0 | 0 |
| 131 | 94127 | 0 | 0 |
| 132 | 94127 | 0 | 0 |
| 133 | 94127 | 0 | 0 |
| 134 | 94127 | 0 | 0 |
| 135 | 94127 | 0 | 0 |
| 136 | 94127 | 0 | 0 |
| 137 | 94127 | 0 | 0 |
| 138 | 94127 | 0 | 0 |
| 139 | 94127 | 0 | 0 |
| 140 | 94127 | 0 | 0 |
| 141 | 94127 | 0 | 0 |
| 142 | 94127 | 0 | 0 |
| 143 | 94127 | 0 | 0 |
| 144 | 94127 | 0 | 0 |
| 145 | 94127 | 0 | 0 |
| 146 | 94127 | 0 | 0 |
| 147 | 94127 | 0 | 0 |
| 148 | 94127 | 0 | 0 |
| 149 | 94127 | 0 | 0 |
| 150 | 94127 | 0 | 0 |
| 151 | 94127 | 0 | 0 |
| 152 | 94127 | 0 | 0 |
| 153 | 94127 | 0 | 0 |
| 154 | 94127 | 0 | 0 |
| 155 | 94127 | 0 | 0 |
| 156 | 94127 | 0 | 0 |
| 157 | 94127 | 0 | 0 |
| 158 | 94127 | 0 | 0 |
| 159 | 94127 | 0 | 0 |
| 160 | 94127 | 0 | 0 |
| 161 | 94127 | 0 | 0 |
| 162 | 94127 | 0 | 0 |
| 163 | 94127 | 0 | 0 |
| 164 | 94127 | 0 | 0 |
| 165 | 94127 | 0 | 0 |
| 166 | 94127 | 0 | 0 |
| 167 | 94127 | 0 | 0 |
| 168 | 94127 | 0 | 0 |
| 169 | 94127 | 0 | 0 |
| 170 | 94127 | 0 | 0 |
| 171 | 94127 | 0 | 0 |
| 172 | 94127 | 0 | 0 |
| 173 | 94127 | 0 | 0 |
| 174 | 94127 | 0 | 0 |
| 175 | 94127 | 0 | 0 |
| 176 | 94127 | 0 | 0 |
| 177 | 94127 | 0 | 0 |
| 178 | 94127 | 0 | 0 |
| 179 | 94127 | 0 | 0 |
| 180 | 94127 | 0 | 0 |
| 181 | 94127 | 0 | 0 |
| 182 | 94127 | 0 | 0 |
| 183 | 94127 | 0 | 0 |
| 184 | 94127 | 0 | 0 |
| 185 | 94127 | 0 | 0 |
| 186 | 94127 | 0 | 0 |
| 187 | 94127 | 0 | 0 |
| 188 | 94127 | 0 | 0 |
| 189 | 94127 | 0 | 0 |
| 190 | 94127 | 0 | 0 |
| 191 | 94127 | 0 | 0 |
| 192 | 94127 | 0 | 0 |
| 193 | 94127 | 0 | 0 |
| 194 | 94127 | 0 | 0 |
| 195 | 94127 | 0 | 0 |
| 196 | 94127 | 0 | 0 |
| 197 | 94127 | 0 | 0 |
| 198 | 94127 | 0 | 0 |
| 199 | 94127 | 0 | 0 |
| 200 | 94127 | 0 | 0 |
+-----+-----------------+-------------+------------+
在分析结果的时候，我想统计模型增加的次数。我想使用 df.where(data.hmm_diff > 0).count() 函数来读取我递增一个状态的次数。但是，增加有时会跨越两个状态（即它会跳过中间状态），所以我需要通过对理想化值进行排序来重新分配 HMM 状态标签，以便最低状态为 0，最高状态为 4，等等。是否有一种根据理想化强度将 hmm_state 标签从任意重新分配到 的方法？
例如，标记为“1”的 hmm_state 位于 hmm_state 3 和 4
看起来你只需要像这样定义一个排序的 HMM 状态：
state_orders = {v: i for i, v in enumerate(sorted(df.hmm_idealized.unique()))}
df['sorted_state'] = df.hmm_idealized.map(state_orders)
然后你可以像你在问题中所做的那样继续，但是在这个专栏上做一个差异，并计算它的跳跃。