Generating prediction raster from Random Forest model using R?
我将随机森林模型拟合到R中来自测试站点的表格数据，现在想使用对应于模型中相同预测变量(例如坡度，海拔，pH)的栅格数据生成显示预测概率值的栅格。
建立了RF模型，以使用不同的环境和地球物理数据预测0/1二进制变量
1 2 3 | #random forest model set.seed(321) rf1 <- randomForest(formula=SITE_NONSITE ~., data=dcc.s.dummy, ntree=500, mtry=10) |
dcc.s.dummy包含以下数据：
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 | str(dcc.s.dummy) 'data.frame': 7899 obs. of 25 variables: $ COST_DIST_ECOTONE : num -0.232 0.176 -0.443 -0.478 -0.305 ... $ COST_DIST_HEA : num -0.233 -0.659 -1.055 -0.999 -0.455 ... $ COST_DIST_MEDSTR : num 0.74388 0.63933 0.55964 0.50768 0.00993 ... $ COST_DIST_RIV_COAST : num 0.59 0.63 0.621 0.639 0.617 ... $ DEM30_ASP_RE_2 : num 0 0 0 0 1 0 0 0 0 0 ... $ DEM30_ASP_RE_3 : num 0 1 0 0 0 0 0 0 1 0 ... $ DEM30_ASP_RE_4 : num 1 0 0 0 0 0 0 1 0 0 ... $ DEM30_ASP_RE_5 : num 0 0 1 1 0 1 1 0 0 1 ... $ DEM30_M : num 0.916 0.72 0.499 0.54 1.114 ... $ DEM30_SLOPE : num 0.2063 0.4631 -0.6445 -0.0512 -0.8235 ... $ LOC_REL_RE : num -0.489 -0.476 -0.476 -0.459 -0.661 ... $ LOC_SD_SLOPE : num -0.118 -0.135 -0.316 -0.367 -0.57 ... $ SSURGO_ESRI_DRAINAGE_RE_2: num 0 0 0 0 0 0 0 0 0 0 ... $ SSURGO_ESRI_DRAINAGE_RE_3: num 1 1 1 1 1 1 1 1 1 1 ... $ SSURGO_ESRI_DRAINAGE_RE_4: num 0 0 0 0 0 0 0 0 0 0 ... $ SSURGO_ESRI_DRAINAGE_RE_5: num 0 0 0 0 0 0 0 0 0 0 ... $ SSURGO_ESRI_DRAINAGE_RE_6: num 0 0 0 0 0 0 0 0 0 0 ... $ SSURGO_ESRI_EROSION_RE_2 : num 0 0 0 0 0 1 1 0 0 1 ... $ SSURGO_ESRI_EROSION_RE_3 : num 1 1 1 0 1 0 0 1 1 0 ... $ SSURGO_ESRI_EROSION_RE_4 : num 0 0 0 0 0 0 0 0 0 0 ... $ SSURGO_ESRI_LOC_DIV : num -0.328 -0.188 -0.157 -0.213 -0.652 ... $ SSURGO_ESRI_NATIVEVEG_2 : num 1 1 1 0 1 0 0 1 1 1 ... $ SSURGO_ESRI_NATIVEVEG_3 : num 0 0 0 0 0 1 1 0 0 0 ... $ SSURGO_PH : num 0.813 0.059 1.529 2.32 -1.298 ... $ SITE_NONSITE : Factor w/ 2 levels"0","1": 2 2 2 2 2 1 1 2 2 2 |
然后，我在整个学习区域中获取与这些相同的预测变量相对应的栅格，并将其合并为栅格堆栈：
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 | #plot model predictions COST_DIST_ECOTONE <- raster("cost_dist_ecotone_s.tif.tif") COST_DIST_HEA <- raster("cost_dist_hea_s.tif.tif") COST_DIST_MEDSTR <- raster("cost_dist_medstr_s.tif.tif") COST_DIST_RIV_COAST <- raster("cost_dist_riv_coast_s.tif.tif") DEM30_ASP_RE_2 <- raster("dem30_asp_rel_2.tif.tif") DEM30_ASP_RE_3 <- raster("dem30_asp_rel_3.tif.tif") DEM30_ASP_RE_4 <- raster("dem30_asp_rel_4.tif.tif") DEM30_ASP_RE_5 <- raster("dem30_asp_rel_5.tif.tif") DEM30_M <- raster("dem30_m_s.tif.tif") DEM30_SLOPE <- raster("dem30_slope_s.tif.tif") LOC_REL_RE <- raster("loc_rel_re_s.tif.tif") LOC_SD_SLOPE <- raster("loc_sd_slope_s.tif.tif") SSURGO_ESRI_DRAINAGE_RE_2 <- raster("SSURGO_ESRI_drainage_reclass_nulfill_2.tif.tif") SSURGO_ESRI_DRAINAGE_RE_3 <- raster("SSURGO_ESRI_drainage_reclass_nulfill_3.tif.tif") SSURGO_ESRI_DRAINAGE_RE_4 <- raster("SSURGO_ESRI_drainage_reclass_nulfill_4.tif.tif") SSURGO_ESRI_DRAINAGE_RE_5 <- raster("SSURGO_ESRI_drainage_reclass_nulfill_5.tif.tif") SSURGO_ESRI_DRAINAGE_RE_6 <- raster("SSURGO_ESRI_drainage_reclass_nulfill_6.tif.tif") SSURGO_ESRI_EROSION_RE_2 <- raster("SSURGO_ESRI_erosion_reclass_nulfilll_2.tif.tif") SSURGO_ESRI_EROSION_RE_3 <- raster("SSURGO_ESRI_erosion_reclass_nulfilll_3.tif.tif") SSURGO_ESRI_EROSION_RE_4 <- raster("SSURGO_ESRI_erosion_reclass_nulfilll_4.tif.tif") SSURGO_ESRI_LOC_DIV <- raster("SSURGO_ESRI_loc_div_s.tif.tif") SSURGO_ESRI_NATIVEVEG_2 <- raster("SSURGO_ESRI_nativeveg_nullfill_2.tif.tif") SSURGO_ESRI_NATIVEVEG_3 <- raster("SSURGO_ESRI_nativeveg_nullfill_3.tif.tif") SSURGO_PH <- raster("SSURGO_pH_nullfill_s.tif.tif") ApPl_stack <- stack(COST_DIST_ECOTONE, COST_DIST_HEA, COST_DIST_MEDSTR, COST_DIST_RIV_COAST, DEM30_ASP_RE_2, DEM30_ASP_RE_3, DEM30_ASP_RE_4, DEM30_ASP_RE_5, DEM30_M, DEM30_SLOPE, LOC_REL_RE, LOC_SD_SLOPE, SSURGO_ESRI_DRAINAGE_RE_2, SSURGO_ESRI_DRAINAGE_RE_3, SSURGO_ESRI_DRAINAGE_RE_4, SSURGO_ESRI_DRAINAGE_RE_5, SSURGO_ESRI_DRAINAGE_RE_6, SSURGO_ESRI_EROSION_RE_2, SSURGO_ESRI_EROSION_RE_3, SSURGO_ESRI_EROSION_RE_4, SSURGO_ESRI_LOC_DIV, SSURGO_ESRI_NATIVEVEG_2, SSURGO_ESRI_NATIVEVEG_3, SSURGO_PH) |
但是，尝试在
1 | ApPl_prob <- raster::predict(rf1, newdata=ApPl_stack, type="prob") |
Error in as.data.frame.default(x[[i]], optional = TRUE) : cannot
coerce class ‘structure("RasterLayer", package ="raster")’ to a
data.frame
转换为数据框并使用它会产生此错误：
1 2 | ApPl_df <- as.data.frame(ApPl_stack, xy=TRUE) ApPl_prob <- raster::predict(rf1, newdata=ApPl_df, type="prob") |
Error in model.frame.default(Terms, newdata, na.action = na.omit) :
object is not a matrix In addition: Warning message: 'newdata' had
658242 rows but variables found have 754 rows
在我的每个预测变量栅格中都有658242个单元和754行，这并不是一个巧合。我在这里想念什么？我觉得其中一个功能正在期望没有得到的数据类型。
"对象名称"与图层名称无关，因此您需要设置它们以匹配用于拟合模型的data.frame中的名称。在大多数工作流程中，您会做类似
1 2 3 | f <- c("cost_dist_ecotone_s.tif.tif","cost_dist_hea_s.tif.tif","cost_dist_medstr_s.tif.tif") s <- stack(f) names(s) <- gsub(".tif.tif","", f) |
然后从RasterStack中提取值以适合您的模型-在这种情况下，名称已经匹配。
但是你犯的主要错误是在这里
1 | ApPl_prob <- raster::predict(rf1, newdata=ApPl_stack, type="prob") |
第一个参数应该是RasterStack：
1 | ApPl_prob <- raster::predict(ApPl_stack, rf1, type="prob") |
或者像在答案中一样使用命名参数
1 | raster::predict(model=rf1, object=ApPl_stack, type="prob") |
在仔细检查了以上代码生成的所有对象的结构之后，我发现了问题。出于任何原因，
添加一个额外的步骤来分配匹配名称可以解决该问题：
1 | names(ApPl_stack) <- c("COST_DIST_ECOTONE","COST_DIST_HEA","COST_DIST_MEDSTR","COST_DIST_RIV_COAST","DEM30_ASP_RE_2","DEM30_ASP_RE_3","DEM30_ASP_RE_4","DEM30_ASP_RE_5","DEM30_M","DEM30_SLOPE","LOC_REL_RE","LOC_SD_SLOPE","SSURGO_ESRI_DRAINAGE_RE_2","SSURGO_ESRI_DRAINAGE_RE_3","SSURGO_ESRI_DRAINAGE_RE_4","SSURGO_ESRI_DRAINAGE_RE_5","SSURGO_ESRI_DRAINAGE_RE_6","SSURGO_ESRI_EROSION_RE_2","SSURGO_ESRI_EROSION_RE_3","SSURGO_ESRI_EROSION_RE_4","SSURGO_ESRI_LOC_DIV","SSURGO_ESRI_NATIVEVEG_2","SSURGO_ESRI_NATIVEVEG_3","SSURGO_PH") |
然后，我可以使用以下代码毫无问题地生成和绘制预测：
1 2 3 4 5 | #plot predictions and save raster to file ApPl_prob <- 1- raster::predict(model=rf1, object=ApPl_stack, type="prob") palette <- matlab.like(20) plot(ApPl_prob, col=palette) writeRaster(ApPl_prob,"ApPl_prob", format='GTiff') |
