IFANN: Intuitionistic Fuzzy Artificial Neural Network for Time...

View source: R/IFANN.R

IFANNR Documentation

Intuitionistic Fuzzy Artificial Neural Network for Time Series

Description

Fits an intuitionistic fuzzy artificial neural network (IFANN) model to a univariate time series. The lagged values of the series and their intuitionistic fuzzy membership and non-membership degrees (from intuitionistic fuzzy c-means clustering) are used as inputs of a feed-forward neural network with one hidden layer. The lag order and the network hyper-parameters can each be set manually or tuned automatically by grid search.

Usage

IFANN(
  Data,
  Class = 3,
  Split = 0.8,
  Lag_Tuning = c("auto", "manual"),
  Lag = 1,
  Lag_Range = c(min = 1, max = 12, step = 1),
  ANN_Tuning = c("auto", "manual"),
  Size = 4,
  Decay = 0.01,
  Size_Range = c(min = 2, max = 8, step = 2),
  Decay_Range = c(min = 0.01, max = 0.09, step = 0.04),
  Maxit = 200,
  Repeats = 5,
  Validation = c("holdout", "tscv"),
  Val_Split = 0.2,
  Folds = 3,
  Metric = c("RMSE", "MAE", "MAPE"),
  Forecast_Type = c("one-step", "recursive"),
  m = 2,
  Alpha = 0.85,
  Max_Iter = 100,
  Tol = 1e-05,
  Verbose = TRUE
)

Arguments

Data

Numeric time series (vector or ts) without missing values.

Class

Number of intuitionistic fuzzy clusters (at least 2).

Split

Proportion of the data used for training, strictly between 0 and 1.

Lag_Tuning

"auto" (default) to tune the lag order over Lag_Range, or "manual" to use Lag.

Lag

Lag order used when Lag_Tuning = "manual" (a single positive integer).

Lag_Range

Lag search range c(min, max, step) for Lag_Tuning = "auto" (positive integers). Lags too large for the data are dropped with a warning.

ANN_Tuning

"auto" (default) to tune Size and Decay, or "manual" to use the single values given.

Size, Decay

Number of hidden neurons (a positive integer) and weight decay (non-negative), used when ANN_Tuning = "manual".

Size_Range, Decay_Range

Search ranges c(min, max, step) for ANN_Tuning = "auto". Size values must be positive integers and Decay values non-negative.

Maxit

Maximum number of training iterations of each network.

Repeats

Number of networks fitted from different random starting weights; their predictions are averaged.

Validation

"holdout" (default) or "tscv" (expanding-window time-series cross-validation).

Val_Split

Share of the training set used for validation, strictly between 0 and 1.

Folds

Number of folds for Validation = "tscv".

Metric

Criterion minimised during tuning: "RMSE" (default), "MAE" or "MAPE".

Forecast_Type

"one-step" (default): every test value is predicted from the observed previous values. "recursive": the test period is forecast multi-step ahead, feeding predictions back as lags. The same type is used during validation.

m

Fuzzifier of the clustering (greater than 1).

Alpha

Exponent of the Yager non-membership generator, in (0, 1].

Max_Iter, Tol

Maximum number of iterations and convergence tolerance of the clustering.

Verbose

Logical; show the grid-search progress and the result summary as messages (default TRUE).

Details

Model. The first round(length(Data) * Split) observations form the training set and the rest the test set. Intuitionistic fuzzy c-means clustering (Chaira, 2011) with Class clusters is fitted to the training observations only. Non-membership degrees come from the Yager-type generator \nu = (1 - \mu^\alpha)^{1/\alpha}, so the hesitation degree \pi = 1 - \mu - \nu is non-negative for 0 < \alpha \le 1; Alpha = 1 gives ordinary fuzzy sets. For a target y_t and lag order L the inputs are y_{t-1}, \dots, y_{t-L} (min-max scaled with training statistics) and the membership and non-membership degrees of each lagged value in every cluster, i.e. p = L(1 + 2\,Class) inputs. Only past values enter the inputs, so the test-set forecasts use no information from the values being forecast.

Network. The network has p inputs, one hidden layer of Size logistic neurons and a linear output, and is fitted to the target scaled to [0, 1] with nnet (BFGS with weight decay Decay, at most Maxit iterations). Because the result depends on the random starting weights, Repeats networks are fitted from different starts and their predictions are averaged. Maxit and Repeats are fixed settings used for every model, including during tuning.

Tuning. Lag_Tuning and ANN_Tuning are set independently:

  • "manual": the single values Lag and / or Size, Decay are used as given.

  • "auto": the parameters are tuned by grid search. Each search range is given as c(min, max, step) and expands to seq(min, max, by = step); for example Lag_Range = c(1, 12, 1) gives the lags 1 to 12 and Size_Range = c(2, 8, 2) gives 2, 4, 6, 8 hidden neurons. If max is not reached exactly by the steps, the last value is the largest one below max. Use min = max for a single value.

When both are "auto" every combination of lag and network parameters is evaluated. With Verbose = TRUE (default) the search prints the values of every parameter, the number of combinations, and a progress bar with the best score so far, the elapsed time and the estimated time remaining.

Validation. Candidates are compared on the last Val_Split share of the training set. With "holdout" the model is fitted on the observations before that block and evaluated on the block. With "tscv" the block is divided into Folds consecutive parts and each part is predicted from a model fitted on all earlier observations (expanding window). The clustering and scaling are re-estimated on each fitting window. After tuning, the model is refitted on the whole training set with the best parameters.

Output. With Verbose = TRUE a summary of the selected parameters and the accuracy is shown as a message at the end (silence it with Verbose = FALSE or suppressMessages()), and the result is returned invisibly; assign it (e.g. Model <- IFANN(...)) to keep it.

Reproducibility. The clustering starts from a random membership matrix and the network weights are initialised at random, so call set.seed() before IFANN() to obtain reproducible results.

Value

A list (returned invisibly) with components

Accuracy

Matrix of RMSE, MAE and MAPE (%) for the training and test sets.

Train_Fitted

Fitted values for the training targets (the first Lag training observations have no fitted value).

Test_Prediction

Predictions for the test set.

Train_Actual, Test_Actual

The corresponding observed values.

Best_Params

List with Lag, Size and Decay of the final model.

Architecture

Number of inputs, hidden neurons and outputs, number of weights per network, Maxit, Repeats and Converged, the number of networks whose optimiser stopped before reaching Maxit.

Tuning

NULL when nothing was tuned; otherwise a list with the validation scheme, metric, number of combinations, best validation score, elapsed time and History (a data frame with the score of every combination).

Networks

List of the Repeats fitted nnet objects of the final model (fitted to the target scaled to [0, 1]).

Scaling

Minimum and maximum of the training data used for the [0, 1] scaling.

IFCM

The intuitionistic fuzzy clustering of the training set (centres, membership, non-membership and hesitation degrees).

Settings

Other settings used.

References

Yolcu, O.C. and Yolcu, U. (2023). A novel intuitionistic fuzzy time series prediction model with cascaded structure for financial time series. Expert Systems with Applications, 215, 119336. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1016/j.eswa.2022.119336")}

Chaira, T. (2011). A novel intuitionistic fuzzy C means clustering algorithm and its application to medical images. Applied Soft Computing, 11(2), 1711–1717.

Venables, W.N. and Ripley, B.D. (2002). Modern Applied Statistics with S. Fourth edition. Springer, New York.

Examples

set.seed(123)
dat <- 50 + 10 * sin(2 * pi * (1:120) / 12) + rnorm(120)

# Everything manual: lag 12, 4 hidden neurons, weight decay 0.05
Model1 <- IFANN(dat, Class = 3, Split = 0.8,
                Lag_Tuning = "manual", Lag = 12,
                ANN_Tuning = "manual", Size = 4, Decay = 0.05,
                Repeats = 3)
Model1$Accuracy

# Manual lag, network parameters tuned on a very small grid
Model2 <- IFANN(dat, Class = 3, Split = 0.8,
                Lag_Tuning = "manual", Lag = 12,
                Size_Range = c(2, 4, 2),           # 2, 4
                Decay_Range = c(0.01, 0.09, 0.08), # 0.01, 0.09
                Repeats = 2, Verbose = FALSE)
Model2$Best_Params


# Lag and network parameters tuned jointly (small grid)
Model3 <- IFANN(dat, Class = 3, Split = 0.8,
                Lag_Range = c(2, 12, 5),           # 2, 7, 12
                Size_Range = c(2, 4, 2),           # 2, 4
                Decay_Range = c(0.05, 0.05, 0),    # 0.05
                Repeats = 2, Verbose = FALSE)
head(Model3$Tuning$History)



IFANN documentation built on Nov. 20, 2026, 5:12 p.m.

Related to IFANN in IFANN...