# Custom R Script for H2O Autoencoder **URL:** \ **Category:** HowTo\ **Created:** [November 20, 2020, 1:21pm UTC](https://community.exploratory.io/t/custom-r-script-for-h2o-autoencoder/1947 "2020-11-20T13:21:41Z")\ **Posts on this page:** 1\ **Page:** 1
**Author:** ![Andrew\_Wilcox](https://community.exploratory.io/user_avatar/community.exploratory.io/andrew_wilcox/32/1041_2.png) [@Andrew\_Wilcox](https://community.exploratory.io/u/Andrew_Wilcox)\ **Post date:** [November 20, 2020, 1:21pm UTC](https://community.exploratory.io/t/custom-r-script-for-h2o-autoencoder/1947/1 "2020-11-20T13:21:41Z")
I am using the Turbofan Engine Degradation Simulation Data Set from NASA located here: > **[NASA Ames Intelligent Systems Division home](https://www.nasa.gov/intelligent-systems-division/)** > > We provide leadership in information technologies by conducting mission-driven, user-centric research and development in computational sciences for NASA applications. We demonstrate and infuse innovative technologies for autonomy, robotics,... I am trying to remove noise from the sensor data by applying the H2O autoencoder, as recommended here: > **[Machine Learning Techniques for Predictive Maintenance](https://www.infoq.com/articles/machine-learning-techniques-predictive-maintenance/)** > > In this article, the authors explore how we can build a machine learning model to do predictive maintenance of systems. They discuss a sample application using NASA engine failure dataset to predict the Remaining Useful Time (RUL) with regression... I have followed these instructions to get H2o installed and setup in Exploratory: [https://blog.exploratory.io/using-h2o-powered-machine-learning-algorithms-in-r-exploratory-9c514fddca8d](https://blog.exploratory.io/using-h2o-powered-machine-learning-algorithms-in-r-exploratory-9c514fddca8d) I created a custom script, which runs, but the data I return is not in the correct format for Exploratory, and I get the resulting error: > Please define tidy.Problem with `mutate()` input `model`. ✖ cannot coerce class ‘structure(H2OAutoEncoderModel, package = h2o)’ ℹ Input `model` is `purrr::map2(...)`.. Below is my script and the output from H2O service. Thanks in advance. > # library(h2o) > > h2o.init(max\_mem\_size=“60g”, nthreads=-1) > > # Utility function that converts regular data.frame into H2OFrame. > > as\_h2o ← function(df) { > > # Convert character columns into factor. > > for (colname in colnames(df)) { > if (class(df[[colname]]) == “character”) { > df[[colname]] ← as.factor(df[[colname]]) > } > } > df ← as.h2o(df) > df > } > > build\_h2o\_autoencoder ← function(data) { > > ``` > i <- 1:200 > h2o_data.train <- data[i, -1] > h2o_data.train <- as_h2o(h2o_data.train) > > itest <- 201:295 > h2o_data.test <- data[itest, -1] > h2o_data.test <- as_h2o(h2o_data.test) > > #xnames <- colnames(h2o_data.train) > xnames = c(4:25) > > h2o_data <- as_h2o(data) # convert data into H2OFrame. > > # Train a denoise autoencoder > denoise_ae <- h2o.deeplearning( > x = xnames, > training_frame = h2o_data.train, > validation_frame = h2o_data.test, > autoencoder = TRUE, > hidden = 3, > activation = 'Tanh', > sparse = TRUE > ) > # convert predicted data back to data.frame. > denoise_ae <- as.data.frame(denoise_ae) > denoise_ae > > ``` > > } Partial output from the H2O service > 11-20 09:07:43.577 192.168.2.16:54321 93144 FJ-1-3 INFO: Finished training the Deep Learning model. > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Model Metrics Type: AutoEncoder > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Description: N/A > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: model id: DeepLearning\_model\_R\_1605876768431\_4 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: frame id: df.temporary.sample.16.49% > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: MSE: 0.13167608 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: RMSE: 0.36287197 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Model Metrics Type: AutoEncoder > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Description: N/A > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: model id: DeepLearning\_model\_R\_1605876768431\_4 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: frame id: df > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: MSE: 0.13092372 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: RMSE: 0.36183384 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Variable Importances: > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Variable Relative Importance Scaled Importance Percentage > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 5.70 1.000000 1.000000 0.078082 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 1.01 0.895548 0.895548 0.069926 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 41.69 0.747241 0.747241 0.058346 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 445.00 0.709292 0.709292 0.055383 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 137.36 0.703748 0.703748 0.054950 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 8311.32 0.687593 0.687593 0.053689 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 100.0 0.667932 0.667932 0.052154 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 3.91 0.643408 0.643408 0.050239 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 1112.93 0.603242 0.603242 0.047103 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 8074.83 0.593626 0.593626 0.046352 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: — > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2211.86 0.508927 0.508927 0.039738 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 100.00 0.482159 0.482159 0.037648 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 549.68 0.439516 0.439516 0.034318 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 10.62 0.429863 0.429863 0.033565 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 0.02 0.411523 0.411523 0.032133 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 129.78 0.407064 0.407064 0.031784 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2212 0.390483 0.390483 0.030490 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 1343.43 0.387120 0.387120 0.030227 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 6.3670 0.272774 0.272774 0.021299 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 1 0.207339 0.207339 0.016190 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Status of Neuron Layers (auto-encoder, gaussian distribution, Quadratic loss, 164 weights/biases, 8.1 KB, 606,360 training samples, mini-batch size 1): > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Layer Units Type Dropout L1 L2 Mean Rate Rate RMS Momentum Mean Weight Weight RMS Mean Bias Bias RMS > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 1 23 Input 0.00 % > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2 3 Tanh 0 0.000000 0.000000 1.004224 0.000000 0.000000 0.016175 0.254712 0.000000 0.000000 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 3 23 Tanh 0.000000 0.000000 1.004224 0.000000 0.000000 -0.022812 0.253095 0.000000 0.000000 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Scoring History: > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: Timestamp Duration Training Speed Epochs Iterations Samples Training RMSE Training MSE Validation RMSE Validation MSE > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2020-11-20 09:07:40 0.294 sec 0.00000 obs/sec 0.00000 0 0.000000 0.36287 0.13168 0.36183 0.13092 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2020-11-20 09:07:43 3.340 sec 210468 obs/sec 10.00000 10 606360.000000 0.78124 0.61034 0.78188 0.61133 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: 2020-11-20 09:07:43 3.442 sec 210249 obs/sec 10.00000 10 606360.000000 0.36287 0.13168 0.36183 0.13092 > 11-20 09:07:43.579 192.168.2.16:54321 93144 FJ-1-3 INFO: ============================================================================================================================================================================== > 11-20 09:07:44.369 192.168.2.16:54321 93144 #22755-14 INFO: GET /3/Models/DeepLearning\_model\_R\_1605876768431\_4, parms: {}