Fake News Detection Technical Models
WHO ❓
WHAT ❓
WHY ❓
Machine Learning-Based Models :learn patterns from labeled data, making predictions without explicit programming. They rely on manual feature extraction, including linguistic, theme, user, and communication features, to improve classification accuracy.
Deep Learning-Based Models : subset of machine learning algorithms that are inspired by the structure and function of the human brain's neural networks. These models are capable of learning intricate patterns from raw data through multiple layers of abstraction. They excel at tasks involving large volumes of unstructured data, such as images, text, and audio.
Pretrained Models : are neural network architectures that have been trained on vast amounts of data to learn generic features or representations. These models leverage transfer learning, where knowledge gained from pretraining on one task is transferred to another related task with limited labeled data.
Machine Learning-Based Models
Deep Learning-Based Models
Pretrained Models
Deep Learning-Based Models
Pretrained Models
Machine Learning-Based Models
Random Forest : operates by constructing multiple decision trees during training and outputting the mode of the classes (classification) or the average prediction (regression) of the individual trees.
Naive Bayes : simple probabilistic classifier based on Bayes theorem with the naive assumption of feature independence.
Support Vector Classifier : aims to maximize the margin between classes while minimizing classification errors, making it effective for both linear and non-linear classification tasks.
Convolutional Neural Networks : class of deep neural networks commonly applied to analyze visual imagery
Bidirectional LSTM : extension of the traditional LSTM architecture that processes input sequences in both forward and backward directions.
Long Short Term Memory : type of recurrent neural network (RNN) architecture designed to overcome the vanishing gradient problem in traditional RNNs.
DistilBERT :distilled version of BERT (smaller in size and faster)
RoBERTa :optimized version of BERT developed by Facebook AI
BERT : transformer-based model developed by Google
AlBERT : a lighter version of BERT
Data scientists, machine learning engineers, and researchers who develop and train models using machine learning algorithms. They are responsible for data preprocessing, feature engineering, model selection, and evaluation.
Deep learning researchers, neural network architects, and data scientists who specialize in developing and training deep learning models. They have expertise in designing neural network architectures, optimizing model parameters, and leveraging computational resources for training large-scale models.
Natural language processing (NLP) researchers, machine learning practitioners, and data scientists who work with pretrained models. They fine-tune and adapt pretrained models to specific tasks, evaluate model performance, and integrate them into applications or systems for tasks such as fake news detection.
Machine learning-based models are used because they are effective for tasks where manual feature engineering is feasible and where interpretability of the model is important. They are well-suited for scenarios where labeled data is available, allowing for the training of models to make predictions based on learned patterns and relationships in the data.
Deep learning-based models are employed because they excel at learning intricate patterns and representations from raw data, particularly in tasks involving unstructured data such as images, text, and audio. They are capable of automatically extracting hierarchical features from data, reducing the need for manual feature engineering and often achieving state-of-the-art performance.
Pretrained models are utilized because they leverage large-scale pretraining on extensive datasets, capturing generic features or representations that can be transferred to downstream tasks with limited labeled data. They enable efficient transfer learning, allowing practitioners to adapt pretrained models to specific tasks, such as fake news detection, with reduced computational resources and training time.