Time Series models for keras in R
You can install the development version of aion from GitHub with:
# install.packages("devtools")
devtools::install_github("krzjoa/aion")- Temporal Fusion Transformer model
- additional layers: Temporal Convolutional Network block & Legendre Memory Unit
make_arrayandts_generatorfunctions to quickly prepare input/output for{keras}time series models- new loss functions:
loss_quantile,loss_tweedieandloss_negative_log_likelihood
# Dataset
library(m5)
# Neural Networks
library(aion)
library(keras)
# Data wrangling
library(dplyr, warn.conflicts=FALSE)
library(data.table, warn.conflicts=FALSE)
library(recipes, warn.conflicts=FALSE)
# ==========================================================================
# PREPARING THE DATA
# ==========================================================================
train <- tiny_m5[date < '2016-01-01']
test <- tiny_m5[date >= '2016-01-01']
m5_recipe <-
recipe(value ~ ., data=train) %>%
step_mutate(item_id_idx=item_id, store_id_idx=store_id) %>%
step_integer(item_id_idx, store_id_idx,
wday, month,
event_name_1, event_type_1,
event_name_2, event_type_2,
zero_based=TRUE) %>%
step_naomit(all_predictors()) %>%
prep()
train <- bake(m5_recipe, train)
test <- bake(m5_recipe, test)
TARGET <- 'value'
STATIC <- c('item_id_idx', 'store_id_idx')
CATEGORICAL <- c('event_name_1', 'event_type_1', STATIC)
NUMERIC <- c('sell_price', 'sell_price')
KEY <- c('item_id', 'store_id')
INDEX <- 'date'
LOOKBACK <- 28
HORIZON <- 14
STRIDE <- LOOKBACK
BATCH_SIZE <- 32
# ==========================================================================
# CREATING GENERATORS
# ==========================================================================
c(train_generator, train_steps) %<-%
ts_generator(
data = train,
key = KEY,
index = INDEX,
lookback = LOOKBACK,
horizon = HORIZON,
stride = STRIDE,
target = TARGET,
static = STATIC,
categorical = CATEGORICAL,
numeric = NUMERIC,
batch_size = BATCH_SIZE
)
c(test_generator, test_steps) %<-%
ts_generator(
data = test,
key = KEY,
index = INDEX,
lookback = LOOKBACK,
horizon = HORIZON,
stride = STRIDE,
target=TARGET,
static=STATIC,
categorical=CATEGORICAL,
numeric=NUMERIC
)The package was initially named aion. However, I was waiting so long with publishing it and somebody has beaten me to it.