从我的随机森林 Model 获得仅响应的问题

Question

this is my first post on Stake overflow so please ask me follow questions if more info is needed!

Situation: I've complied water chemistry data for the Maritimes (Atlantic Canada) for freshwater ecosystems because I am trying to create a predictive species distribution model using a random forest model (RFM) for an invasive species. Unfortunately, Atlantic Canada lack consistent water monitoring programs and the ones that do exist don't monitor for the same parameters as other group. So, my databases (both the training and testing) have many NAs.

Issue: This is the response I keep getting from my RFM:

> p1 <- predict(model2, newdata=Test_Dataset,type="prob")[,2]
> p1

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 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 35 36 37 NA NA NA

What I have tried:

1. I built the RFM (ie model2) using various predictors. I did include:

   model2 <- randomForest(CMS ~ Lat + Lon + pH + Alkalinity + Ca + Hardness + DO + TOC + T_P + T_N + Cond + Na + No_Stocking + No_Fish_Species + Dist_Hwy + No_Boat_Launches + Connected_Lakes + Invasives, importance = TRUE, data=TrainSet, na.action=na.roughfix) model2

**Note that the big list of variables are the predictors and CMS is the species.

2. I tried matching the test dataset (Test_Dataset) with the training dataset (Validation_Dataset).

   Test_Dataset <- rbind(Validation_Dataset[1, ], Validation_Dataset) Test_Dataset <- Test_Dataset[-1,]

3. I have searched for and read multiple resources (including the obvious R pages and references linked there).

4. I have mutated the dataframe as follows (I'll just show the Validation_Dataset as it is the same mutations for both):

   Mutate dataset to fix issues with R reading NA cells

   Validation_Dataset <- Validation_Dataset %>% dplyr::mutate( # convert year into a categorical variable Year = factor (Year), #convert Chlorophyll concentrations from a character file to a number file # convert "NA" into a missing value data whenever appropriate Chlorophyll = dplyr::na_if(Chlorophyll, "NA"), Chlorophyll = factor (Chlorophyll), Hardness = dplyr::na_if(Hardness, "NA"), Hardness= factor (Hardness), Alkalinity = dplyr::na_if(Alkalinity, "NA"), Alkalinity = factor (Alkalinity), Ca = dplyr::na_if(Ca, "NA"), Ca = factor (Ca), TOC = dplyr::na_if(TOC, "NA"), TOC = factor (TOC), Cond = dplyr::na_if(Cond, "NA"), Cond = factor (Cond), Na = dplyr::na_if(Na, "NA"), Na = factor (Cond), NH4 = dplyr::na_if(NH4, "NA"), NH4 = factor (NH4), NO3 = dplyr::na_if(NO3, "NA"), NO3 = factor (NO3), pH = dplyr::na_if(pH, "NA"), pH = factor (pH), T_N = dplyr::na_if(T_N, "NA"), T_N = factor (T_N), T_P = dplyr::na_if(T_P, "NA"), T_P = factor (T_P), DO = dplyr::na_if(DO, "NA"), DO = factor (DO), Salinity = dplyr::na_if(Salinity, "NA"), Salinity = factor (Salinity), No_Stocking = dplyr::na_if(No_Stocking, "NA"), No_Stocking = factor (No_Stocking), No_Fish_Species = dplyr::na_if(No_Fish_Species, "NA"), No_Fish_Species = factor (No_Fish_Species), Dist_Hwy = dplyr::na_if(Dist_Hwy, "NA"), Dist_Hwy = factor (Dist_Hwy), No_Boat_Launches = dplyr::na_if(No_Boat_Launches, "NA"), No_Boat_Launches = factor (No_Boat_Launches), Connected_Lakes = Z15E9FB2F4 0B9E33252CE5FBAF7D8B068Z::na_if(Connected_Lakes, "NA"), Connected_Lakes = factor (Connected_Lakes), Invasives = dplyr::na_if(Invasives, "NA"), Invasives = factor (Invasives), Lat = factor (Lat), Lon = factor (Lon), CMS = factor (CMS))

Question: Does anybody know how to actually make the coding work so that model2 predicts on Test_Dataset? I think this problem might actually be very small, but I'm not seeing it.

Here is a glimpse of the training dataset (Validation_Dataset):

> str(Validation_Dataset)
Classes ‘spec_tbl_df’, ‘tbl_df’, ‘tbl’ and 'data.frame':    37 obs. of  31 variables:
 $ Name            : chr  "Canard River" "Cedar Creek" "Holland River" "Speed River" ...
 $ STN #/COUNTY    : chr  "10000200202" "16001800202" "3007700202" "16018403402" ...
 $ Province        : chr  "ON" "ON" "ON" "ON" ...
 $ Lat             : Factor w/ 37 levels "42.03204214",..: 2 1 11 9 10 8 7 5 6 3 ...
 $ Lon             : Factor w/ 37 levels "-83.01879548",..: 1 2 11 8 10 6 7 9 5 4 ...
 $ Year            : Factor w/ 9 levels "2007, 2011","2010, 2015, 2011",..: 8 8 8 8 8 8 8 8 8 8 ...
 $ Month           : chr  "4" "4" "4" "4" ...
 $ Day             : chr  "11" "12" "26" "27" ...
 $ Data Source     : chr  "ON Provincial (Streams) Water Quality Monitoring Network" "ON Provincial (Streams) Water Quality Monitoring Network" "ON Provincial (Streams) Water Quality Monitoring Network" "ON Provincial (Streams) Water Quality Monitoring Network" ...
 $ pH              : Factor w/ 35 levels "6.073333","6.13",..: 18 21 28 29 25 34 30 32 19 26 ...
 $ Alkalinity      : Factor w/ 31 levels "1.8","2.8","3.933333333",..: 19 22 31 30 27 NA NA 26 NA 21 ...
 $ Hardness        : Factor w/ 13 levels "14.8","36.8",..: 7 8 11 10 9 NA NA 13 NA NA ...
 $ Ca              : Factor w/ 24 levels "3.833333333",..: 18 19 24 20 21 NA NA 22 NA NA ...
 $ Chlorophyll     : Factor w/ 15 levels "0.423601","0.453791",..: NA NA NA NA NA NA NA NA NA NA ...
 $ DO              : Factor w/ 26 levels "0.27","6.2","6.96",..: 21 24 18 16 4 25 17 14 2 7 ...
 $ TOC             : Factor w/ 3 levels "4.8","5.5","8.8": NA NA NA NA NA NA NA NA NA NA ...
 $ T_P             : Factor w/ 24 levels "0.002","0.003",..: 23 22 18 10 15 14 16 13 21 20 ...
 $ T_N             : Factor w/ 32 levels "0.006","0.13",..: 30 31 27 28 17 29 24 32 21 25 ...
 $ NO3+NO2         : num  2.173 2.292 1.092 1.695 0.426 ...
 $ NO3             : Factor w/ 32 levels "0.027","0.035",..: 30 31 26 27 11 29 24 32 22 8 ...
 $ NH4             : Factor w/ 27 levels "0.005","0.006",..: 26 25 22 17 9 11 13 19 23 27 ...
 $ Cond            : Factor w/ 34 levels "41","97","134",..: 24 21 29 23 22 14 34 31 21 17 ...
 $ Salinity        : Factor w/ 9 levels "0.11","0.15",..: NA NA NA NA NA NA NA NA NA NA ...
 $ Na              : Factor w/ 34 levels "41","97","134",..: 24 21 29 23 22 14 34 31 21 17 ...
 $ No_Stocking     : Factor w/ 3 levels "0","1","2": 1 2 2 3 1 2 1 2 1 2 ...
 $ No_Fish_Species : Factor w/ 9 levels "0","1","2","3",..: 1 4 6 4 1 5 1 9 1 9 ...
 $ Dist_Hwy        : Factor w/ 16 levels "0.003","0.006",..: NA NA 16 NA NA NA NA 8 NA 5 ...
 $ No_Boat_Launches: Factor w/ 8 levels "0","1","2","3",..: 1 1 5 1 1 1 1 8 1 3 ...
 $ Connected_Lakes : Factor w/ 11 levels "0","1","2","3",..: 7 2 3 4 9 6 2 3 2 5 ...
 $ Invasives       : Factor w/ 3 levels "0","1","2": NA NA NA NA NA NA NA NA NA NA ...
 $ CMS             : Factor w/ 2 levels "NO","YES": 2 2 2 2 2 2 2 2 2 2 ...

Answer 1

To use the argument na.roughfix. It must first be specified outside the randomForest function if it were to be used. I will use the iris dataset as an example.

iris.roughfix <- na.roughfix(iris.na)
iris.narf <- randomForest(Species ~ ., iris.na, na.action=na.roughfix)