IntelELM is an open-source Python library providing a framework for training Extreme Learning Machines (ELM) using Metaheuristic Algorithms. It is compatible with Scikit-Learn, enabling easy integration into existing machine learning pipelines such as hyperparameter tuning, feature selection,...

🚀 Features

• Free software: GNU General Public License (GPL) V3 license
• Provided Estimator: ElmRegressor, ElmClassifier, MhaElmRegressor, MhaElmClassifier, AutomatedMhaElmTuner, AutomatedMhaElmComparator
• Total Optimization-based ELM Regression: > 200 Models
• Total Optimization-based ELM Classification: > 200 Models
• Supported datasets: 54 (47 classifications and 7 regressions)
• Supported performance metrics: >= 67 (47 regressions and 20 classifications)
• Supported objective functions (as fitness functions or loss functions): >= 67 (47 regressions and 20 classifications)
• Documentation: https://intelelm.readthedocs.io/
• Python versions: >= 3.8.x
• Dependencies: numpy, scipy, scikit-learn, pandas, mealpy, permetrics

📄 Citation Request

If you want to understand how Metaheuristic is applied to Extreme Learning Machine, you need to read the paper titled "A new workload prediction model using extreme learning machine and enhanced tug of war optimization". The paper can be accessed at the following this link

Please include these citations if you plan to use this library:

```bibtex @article{van2025intelelm, title={IntelELM: A python framework for intelligent metaheuristic-based extreme learning machine}, author={Van Thieu, Nguyen and Houssein, Essam H and Oliva, Diego and Hung, Nguyen Duy}, journal={Neurocomputing}, volume={618}, pages={129062}, year={2025}, publisher={Elsevier}, doi={10.1016/j.neucom.2024.129062} }

@article{nguyen2020new, title={A new workload prediction model using extreme learning machine and enhanced tug of war optimization}, author={Nguyen, Thieu and Hoang, Bao and Nguyen, Giang and Nguyen, Binh Minh}, journal={Procedia Computer Science}, volume={170}, pages={362--369}, year={2020}, publisher={Elsevier}, doi={10.1016/j.procs.2020.03.063} }

@article{van2023mealpy, title={MEALPY: An open-source library for latest meta-heuristic algorithms in Python}, author={Van Thieu, Nguyen and Mirjalili, Seyedali}, journal={Journal of Systems Architecture}, year={2023}, publisher={Elsevier}, doi={10.1016/j.sysarc.2023.102871} } ```

📦 Installation

Installation

Install the latest version from PyPI:

bash $ pip install intelelm

Check installed version:

```bash $ python

import intelelm intelelm.version ```

📚 Documentation & Tutorials

• Documentation: https://intelelm.readthedocs.io/en/latest/
• Tutorials:
  • Handwritten Digits Classification
  • California Housing Price Regression

🧪 Example Usage

• In this section, we will explore the usage of the IntelELM model with the assistance of a dataset. While all the preprocessing steps mentioned below can be replicated using Scikit-Learn, we have implemented some utility functions to provide users with convenience and faster usage.

```python

Step 1: Importing the libraries

from intelelm import ElmRegressor, ElmClassifier, MhaElmRegressor, MhaElmClassifier, get_dataset

Step 2: Reading the dataset

data = get_dataset("aniso")

Step 3: Next, split dataset into train and test set

data.splittraintest(testsize=0.2, shuffle=True, randomstate=100)

Step 4: Feature Scaling

data.Xtrain, scalerX = data.scale(data.Xtrain, scalingmethods=("standard", "minmax")) data.Xtest = scalerX.transform(data.X_test)

data.ytrain, scalery = data.encodelabel(data.ytrain) # This is for classification problem only data.ytest = scalery.transform(data.y_test)

Step 5: Fitting ELM-based model to the dataset

5.1: Use standard ELM model for regression problem

regressor = ElmRegressor(layersizes=(10, ), actname="relu", seed=42) regressor.fit(data.Xtrain, data.ytrain)

5.2: Use standard ELM model for classification problem

classifer = ElmClassifier(layersizes=(10, ), actname="tanh", seed=42) classifer.fit(data.Xtrain, data.ytrain)

5.3: Use Metaheuristic-based ELM model for regression problem

print(MhaElmClassifier.SUPPORTEDOPTIMIZERS) print(MhaElmClassifier.SUPPORTEDREGOBJECTIVES) optparas = {"name": "GA", "epoch": 10, "popsize": 30} regressor = MhaElmRegressor(layersizes=(10, ), actname="elu", objname="RMSE", optim="BaseGA", optimparams=optparas, seed=42, lb=None, ub=None, mode='single', nworkers=None, termination=None) regressor.fit(data.Xtrain, data.y_train)

5.4: Use Metaheuristic-based ELM model for classification problem

print(MhaElmClassifier.SUPPORTEDOPTIMIZERS) print(MhaElmClassifier.SUPPORTEDCLSOBJECTIVES) optparas = {"name": "GA", "epoch": 10, "popsize": 30} classifier = MhaElmClassifier(layersizes=(10, ), actname="elu", objname="KLDL", optim="BaseGA", optimparams=optparas, seed=42, lb=None, ub=None, mode='single', nworkers=None, termination=None) classifier.fit(data.Xtrain, data.y_train)

Step 6: Predicting a new result

ypred = regressor.predict(data.Xtest)

ypredcls = classifier.predict(data.Xtest) ypredlabel = scalery.inversetransform(ypred_cls)

Step 7: Calculate metrics using score or scores functions.

print("Try my AS metric with score function") print(regressor.score(data.Xtest, data.ytest, method="AS"))

print("Try my multiple metrics with scores function") print(classifier.scores(data.Xtest, data.ytest, list_methods=["AS", "PS", "F1S", "CEL", "BSL"]))

print("Try my evaluate functions") print(regressor.evaluate(data.ytest, ypred, list_metrics=("RMSE", "MAE", "MAPE", "NSE", "R2")))

Save results

regressor.savelosstrain(savepath="history", filename="losstrain.csv") regressor.savemetrics(data.ytest, ypred, listmetrics=("R2", "MAPE", "MAE", "MSE"), save_path="history", filename="metrics.csv") ```

A real-world dataset contains features that vary in magnitudes, units, and range. We would suggest performing normalization when the scale of a feature is irrelevant or misleading. Feature Scaling basically helps to normalize the data within a particular range.

❓ FAQ

1. How to list supported objective metrics?

Where do I find the supported metrics like above ["AS", "PS", "RS"]. What is that? You can find it here: https://github.com/thieu1995/permetrics or use this

```python from intelelm import MhaElmClassifier, MhaElmRegressor

print(MhaElmRegressor.SUPPORTEDREGOBJECTIVES) print(MhaElmClassifier.SUPPORTEDCLSOBJECTIVES) ```

2. ValueError: Existed at least one new label in y_pred?

I got this type of error python raise ValueError("Existed at least one new label in y_pred.") ValueError: Existed at least one new label in y_pred. How to solve this?

• This occurs only when you are working on a classification problem with a small dataset that has many classes. For instance, the "Zoo" dataset contains only 101 samples, but it has 7 classes. If you split the dataset into a training and testing set with a ratio of around 80% - 20%, there is a chance that one or more classes may appear in the testing set but not in the training set. As a result, when you calculate the performance metrics, you may encounter this error. You cannot predict or assign new data to a new label because you have no knowledge about the new label. There are several solutions to this problem.

- 1st: Use SMOTE to rebalance the dataset:

Use the SMOTE method to address imbalanced data and ensure that all classes have the same number of samples.

```python import pandas as pd from imblearn.over_sampling import SMOTE from intelelm import Data

dataset = pd.readcsv('examples/dataset.csv', indexcol=0).values X, y = dataset[:, 0:-1], dataset[:, -1]

Xnew, ynew = SMOTE().fitresample(X, y) data = Data(Xnew, y_new) ```

• 2st: Try changing random_state in splittraintest:

• Use different randomstate numbers in splittrain_test() function.

```python import pandas as pd from intelelm import Data

dataset = pd.readcsv('examples/dataset.csv', indexcol=0).values X, y = dataset[:, 0:-1], dataset[:, -1] data = Data(X, y) data.splittraintest(testsize=0.2, randomstate=10) # Try different random_state value ```

3. Why don't MHA-based ELM models improve results?

When testing several algorithms based on Extreme Learning Machines (ELM), they all produce the same results. Even during the training process, the global best solution remains unchanged.

• This issue was identified in version <= v1.0.2 when the default values for the lower bound (lb) and upper bound (ub) were set in the narrow range of (-1, 1). This limited range proved to be too small, causing all algorithms to converge to local optima. Fortunately, this problem has been addressed in versions > v1.0.3, where the default range has been extended to (-10., 10.). You also can define your own lb and ub ranges depend on your problem.
• In traditional neural network like MLP, they weights (weights + biases) are typically initialized within the range of (-1., 1.). However, during training using gradient-based methods, these values are updated, and there are no strict bounds on them.
• Meanwhile, in metaheuristic optimization, it's necessary to set boundaries for decision variables (weights) each time a new search agent is formed. Therefore, if you define a narrow range, your optimizer may converge more quickly, but it's more likely to get stuck in local optima (which explains why the global best value remains unchanged during training). Moreover, in some cases, there might not even be a global optimum within that narrow range. Conversely, if you set a wider range, the optimization process may be slower, and the global best value may change more gradually. In such cases, you might need to increase the number of epochs, perhaps up to 1000, for the optimizer to explore the solution space thoroughly.

```python from intelelm import MhaElmClassifier

optparas = {"name": "GA", "epoch": 30, "popsize": 30} model = MhaElmClassifier(layersizes=(10, ), actname="elu", objname="KLDL", optim="BaseGA", optimparams=optparas, verbose=True, seed=42, lb=-10., ub=10., mode='single', nworkers=None, termination=None) model.fit(Xtrain, ytrain) ypred = model.predict(Xtest) ```

🔗 Useful Links

• 📦 Source Code
• 📖 Documentation
• ⬇️ PyPI Releases
• ❗ Report Issues
• 📝 Changelog
• 💬 Chat Group

🤝 Related Projects

• MEALPY
• Metaheuristics
• Opfunu
• Enoppy
• Permetrics
• MetaCluster
• Pfevaluator
• AIIR Team

Developed by: Thieu @ 2023