我正在尝试使用泰坦尼克号数据集使用keras来预测生存，但我不断收到错误消息。我无法弄清原因：（
> str(full_data)
Classes ‘grouped_df’, ‘tbl_df’, ‘tbl’ and 'data.frame': 1309 obs. of 13 variables:
$ survived : num 0 1 1 1 0 0 0 0 1 1 ...
$ pclass : Factor w/ 3 levels "1","2","3": 3 1 3 1 3 3 1 3 3 2 ...
$ sex : Factor w/ 2 levels "female","male": 2 1 1 1 2 2 2 2 1 1 ...
$ ticket : Factor w/ 51 levels "A. 2. ","A./5. ",..: 6 23 45 21 21 21 21 21 21 21 ...
$ cabin : Factor w/ 9 levels "A","B","C","D",..: 8 3 8 3 8 8 5 8 8 8 ...
$ embarked : Factor w/ 4 levels "","C","Q","S": 4 2 4 4 4 3 4 4 4 2 ...
$ fare_factor : Factor w/ 5 levels "1","2","3","4",..: 1 2 1 4 1 1 3 3 3 3 ...
$ age_interval : Factor w/ 6 levels "1","2","3","4",..: 2 4 3 4 4 5 6 1 3 2 ...
$ percent_survive_age : Factor w/ 12 levels "0.0888888888888889",..: 2 11 10 11 3 5 1 7 10 8 ...
$ title : Factor w/ 6 levels "Lady","Miss",..: 3 4 2 4 3 3 3 5 4 4 ...
$ percent_survive_title: Factor w/ 6 levels "0.156673114119923",..: 1 5 4 5 1 1 1 3 5 5 ...
$ fam_size : Factor w/ 9 levels "1","2","3","4",..: 2 2 1 2 1 1 1 5 3 2 ...
$ fam_ID : Factor w/ 176 levels "Abbott_3","Abelson_2",..: 19 40 109 61 109 109 109 130 90 123 ...
- attr(*, "vars")= chr "fam_ID"
- attr(*, "labels")='data.frame': 228 obs. of 1 variable:
..$ fam_ID: chr "Abbott_3" "Abelson_2" "Ahlin_2" "Aks_2" ...
..- attr(*, "vars")= chr "fam_ID"
..- attr(*, "drop")= logi TRUE
- attr(*, "indices")=List of 228
..$ : int 279 746 1283
..$ : int 308 874
..$ : int 40
..$ : int 855 1198
..$ : int 297 305 498 1197
..$ : int 192
..$ : int 13 68 119 541 542 610 813 850 1105
..$ : int 275
..$ : int 518 1081
..$ : int 571
..$ : int 49 353
..$ : int 25 182 233 261 1045 1065 1270
..$ : int 700 1093
..$ : int 206
..$ : int 85
..$ : int 448 469 644 858
..$ : int 362 702
..$ : int 118 299
..$ : int 543 546
..$ : int 183 618 1069 1217
..$ : int 248 871
..$ : int 291 484
..$ : int 140 852 971
..$ : int 188 593 657
..$ : int 356
..$ : int 0 477
..$ : int 670 684 1066 1247
..$ : int 728 1166
..$ : int 78 323 898
..$ : int 578 1257
..$ : int 679 1234
..$ : int 249 854
..$ : int 390 435 763 802
..$ : int 741 987
..$ : int 92 905
..$ : int 724 809
..$ : int 594 1010
..$ : int 166
..$ : int 580 1132
..$ : int 73 1006
..$ : int 1143 1163
..$ : int 426 1219
..$ : int 237 637 801
..$ : int 835 1070 1072
..$ : int 968
..$ : int 348 489 940
..$ : int 160 1016
..$ : int 540 745 1196
..$ : int 1 1125
..$ : int 423 616 1092
..$ : int 671
..$ : int 983
..$ : int 549 565 900 1078 1221
..$ : int 347 949
..$ : int 559 1151
..$ : int 93 788 923 1245
..$ : int 361 906
..$ : int 690 781
..$ : int 445 1184 1265
..$ : int 98 651
..$ : int 544 1130
..$ : int 1075
..$ : int 416 1085 1138
..$ : int 556 599
..$ : int 866 1111
..$ : int 981 1063
..$ : int 1041
..$ : int 352 532 1228
..$ : int 496
..$ : int 53 1168
..$ : int 86 147 436 736 1058
..$ : int 27 88 341 438 944 960
..$ : int 334 1295
..$ : int 660
..$ : int 587 1288
..$ : int 539
..$ : int 3 137
..$ : int 405 1292
..$ : int 1259 1293
..$ : int 861 1261
..$ : int 453 849
..$ : int 165 328 548
..$ : int 59 71 386 480 678 683 1030 1031
..$ : int 268 332
..$ : int 97 1241
..$ : int 104 392
..$ : int 451 490
..$ : int 142 403
..$ : int 247 755
..$ : int 704 1200
..$ : int 860
..$ : int 370 1255
..$ : int 52 645 720 848
..$ : int 62 230
..$ : int 314 440 535
..$ : int 820 1199
..$ : int 615 754 1244 1276
..$ : int 120 655 665
..$ : int 1129
.. [list output truncated]
- attr(*, "drop")= logi TRUE
- attr(*, "group_sizes")= int 3 2 1 2 4 1 9 1 2 1 ...
- attr(*, "biggest_group_size")= int 790
> full_data_dummy <- as.matrix(createDummyFeatures(full_data[1:891,]))
> str(full_data_dummy)
num [1:891, 1:290] 0 1 1 1 0 0 0 0 1 1 ...
- attr(*, "dimnames")=List of 2
..$ : chr [1:891] "1" "2" "3" "4" ...
..$ : chr [1:290] "survived" "pclass.1" "pclass.2" "pclass.3" ...
>
> ## Using keras for deeplearning classification
>
> library(keras)
> ## defining a network with 2 hidden layers with 100 units, using a relu activation in order to initally overfit model
> model <- keras_model_sequential() %>%
+ layer_dense(units = 16, activation = "relu", input_shape = c(289)) %>%
+ layer_dense(units = 16, activation = "relu") %>%
+ layer_dense(units = 1, activation = "sigmoid")
>
> ## spliting train data into train and validation, in order to monitor how much model overfits the train data
> val_indices <- 1:300
> x_val <- full_data_dummy[val_indices,-1]
> x_partial <- full_data_dummy[-val_indices,-1]
> y_val <- full_data_dummy[val_indices,1]
> y_partial <- full_data_dummy[-val_indices,1]
>
> ## compiling model using binary crossenthropy and metrics accuracy
> model %>% compile(
+ optimizer = "rmsprop",
+ loss = "binary_crossentropy",
+ metrics = c("accuracy"))
>
> ## fitting model on training dataset
> history <- model %>% fit(
+ x_partial, y_partial, epochs = 20, batch_size = 50, validation_data = c(x_val,y_val))
Error in py_call_impl(callable, dots$args, dots$keywords) :
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
从示例代码中可以看到，我将所有因子变量都设置为虚拟变量，并将虚拟数据帧转换为矩阵。然后我拆分了火车和测试数据。我已经包括了完整数据帧和虚拟数据帧的结构。