I'm trying to find patterns in a large dataset using the neuralnet package.
My data file looks something like this (30,204,447 rows) :
id.company,EPS.or.Sales,FQ.or.FY,fiscal,date,value
000001,EPS,FY,2001,20020201,-5.520000
000001,SAL,FQ,2000,20020401,70.300003
000001,SAL,FY,2001,20020325,49.200001
000002,EPS,FQ,2008,20071009,-4.000000
000002,SAL,FY,2008,20071009,1.400000
I have split this initial file into four new files for annual/quarterly sales/EPS and it is on those files that I want to use neural networks to see if I can use the variables id.company, fiscal and date in the case below to predict the annual sales results.
To do so, I have written the following code:
dataset <- read.table("fy_sal_data.txt",header=T, sep="\t") #my file doesn't actually use comas as separators
#extract training set and testing set
trainset <- dataset[1:1000, ]
testset <- dataset[1001:2000, ]
#building the NN
ann <- neuralnet(value ~ id.company + fiscal + date, trainset, hidden = 3,
lifesign="minimal", threshold=0.01)
#testing the output
temp_test <- subset(testset, select=c("id.company", "fiscal", "date"))
ann.results <- compute(ann, temp_test)
#display the results
cleanoutput <- cbind(testset$value, as.data.frame(ann.results$net.result))
colnames(cleanoutput) <- c("Expected Output", "NN Output")
head(cleanoutput, 30)
Now my problem is that the compute function returns a constant answer no matter the inputs of the testing set.
Expected Output NN Output
1001 2006.500000 1417.796651
1002 2009.000000 1417.796651
1003 2006.500000 1417.796651
1004 2002.500000 1417.796651
I am very new to R and its neural networks packages but I have found online that some of the reasons for such results can be either:
an insufficient number of training examples (here I'm using a thousand ones but I've also tried using a million rows and the results were the same, only it took 4h to train)
or an error in the formula.
I am sure I'm doing something wrong but I can't seem to figure out what.
How can I perform an operation (like subsetting or adding a calculated column) on each imputed dataset in an object of class mids from R's package mice? I would like the result to still be a mids object.
Edit: Example
library(mice)
data(nhanes)
# create imputed datasets
imput = mice(nhanes)
The imputed datasets are stored as a list of lists
imput$imp
where there are rows only for the observations with imputation for the given variable.
The original (incomplete) dataset is stored here:
imput$data
For example, how would I create a new variable calculated as chl/2 in each of the imputed datasets, yielding a new mids object?
This can be done easily as follows -
Use complete() to convert a mids object to a long-format data.frame:
long1 <- complete(midsobj1, action='long', include=TRUE)
Perform whatever manipulations needed:
long1$new.var <- long1$chl/2
long2 <- subset(long1, age >= 5)
use as.mids() to convert back manipulated data to mids object:
midsobj2 <- as.mids(long2)
Now you can use midsobj2 as required. Note that the include=TRUE (used to include the original data with missing values) is needed for as.mids() to compress the long-formatted data properly. Note that prior to mice v2.25 there was a bug in the as.mids() function (see this post https://stats.stackexchange.com/a/158327/69413)
EDIT: According to this answer https://stackoverflow.com/a/34859264/4269699 (from what is essentially a duplicate question) you can also edit the mids object directly by accessing $data and $imp. So for example
midsobj2<-midsobj1
midsobj2$data$new.var <- midsobj2$data$chl/2
midsobj2$imp$new.var <- midsobj2$imp$chl/2
You will run into trouble though if you want to subset $imp or if you want to use $call, so I wouldn't recommend this solution in general.
Another option is to calculate the variables before the imputation and place restrictions on them.
library(mice)
# Create the additional variable - this will have missing
nhanes$extra <- nhanes$chl / 2
# Change the method of imputation for extra, so that it always equals chl/2
# Change the predictor matrix so only chl predicts extra
ini <- mice(nhanes, max = 0, print = FALSE)
meth <- ini$meth
meth["extra"] <- "~I(chl / 2)"
pred <- ini$pred # extra isn't used to predict
pred["extra", "chl"] <- 1
# Imputations
imput <- mice(nhanes, seed = 1, pred = pred, meth = meth, print = FALSE)
There are examples in mice: Multivariate Imputation by Chained Equations in R.
There is an overload of with that can help you here
with(imput, chl/2)
the documentation is given at ?with.mids
There's a function for this in the basecamb package:
library(basecamb)
apply_function_to_imputed_data(mids_object, function)
I'm having a similar problem to the questioners here had with the linear model predict function, but I am trying to use the "time series linear model" function from Rob Hyndman's forecasting package.
Predict.lm in R fails to recognize newdata
predict.lm with newdata
totalConv <- ts(varData[,43])
metaSearch <- ts(varData[,45])
PPCBrand <- ts(varData[,38])
PPCGeneric <- ts(varData[,34])
PPCLocation <- ts(varData[,35])
brandDisplay <- ts(varData[,29])
standardDisplay <- ts(varData[,3])
TV <- ts(varData[,2])
richMedia <- ts(varData[,46])
df.HA <- data.frame(totalConv, metaSearch,
PPCBrand, PPCGeneric, PPCLocation,
brandDisplay, standardDisplay,
TV, richMedia)
As you can see I've tried to avoid the names issues by creating a data frame of the time series objects.
However, I then fit a tslm object (time series linear model) as follows -
fit1 <- tslm(totalConv ~ metaSearch
+ PPCBrand + PPCGeneric + PPCLocation
+ brandDisplay + standardDisplay
+ TV + richMedia data = df.HA
)
Despite having created a data frame and named all the objects properly I get the same dimension error as these other users have experienced.
Error in forecast.lm(fit1) : Variables not found in newdata
In addition: Warning messages:
1: 'newdata' had 10 rows but variables found have 696 rows
2: 'newdata' had 10 rows but variables found have 696 rows
the model frame seems to give sensible names to all of the variables, so I don't know what is up with the forecast function:-
names(model.frame(fit1))
[1] "totalConv" "metaSearch" "PPCBrand" "PPCGeneric" "PPCLocation" "brandDisplay"
[7] "standardDisplay" "TV" "richMedia"
Can anyone suggest any other improvements to my model specification that might help the forecast function to run?
EDIT 1: Ok, just so there's a working example, I've used the data given in Irsal's answer to this question (converting to time series objects) and then fitted the tslm. I get the same error (different dimensions obviously):-
Is there an easy way to revert a forecast back into a time series for plotting?
I'm really confused about what I'm doing wrong, my code looks identical to that used in all of the examples on this....
data <- c(11,53,50,53,57,69,70,65,64,66,66,64,61,65,69,61,67,71,74,71,77,75,85,88,95,
93,96,89,95,98,110,134,127,132,107,94,79,72,68,72,70,66,62,62,60,59,61,67,
74,87,112,134,51,50,38,40,44,54,52,51,48,50,49,49,48,57,52,53,50,50,55,50,
55,60,65,67,75,66,65,65,69,72,93,137,125,110,93,72,61,55,51,52,50,46,46,45,
48,44,45,53,55,65,89,112,38,7,39,35,37,41,51,53,57,52,57,51,52,49,48,48,51,
54,48,50,50,53,56,64,71,74,66,69,71,75,84,93,107,111,112,90,75,62,53,51,52,
51,49,48,49,52,50,50,59,58,69,95,148,49,83,40,40,40,53,57,54,52,56,53,55,
55,51,54,45,49,46,52,49,50,57,58,63,73,66,63,72,72,71,77,105,97,104,85,73,
66,55,52,50,52,48,48,46,48,53,49,58,56,72,84,124,76,4,40,39,36,38,48,55,49,
51,48,46,46,47,44,44,45,43,48,46,45,50,50,56,62,53,62,63)
data2 <- c(rnorm(237))
library(forecast)
nData <- ts(data)
nData2 <- ts(data2)
dat.ts <- tslm(nData~nData2)
forecast(dat.ts)
Error in forecast.lm(dat.ts) : Variables not found in newdata
In addition: Warning messages:
1: 'newdata' had 10 rows but variables found have 237 rows
2: 'newdata' had 10 rows but variables found have 237 rows
EDIT 2: Same error even if I combine both series into a data frame.
nData.df <- data.frame(nData, nData2)
dat.ts <- tslm(nData~nData2, data = nData.df)
forecast(dat.ts)
tslm fits a linear regression model. You need to provide the future values of the explanatory variables if you want to forecast. These should be provided via the newdata argument of forecast.lm.