The CM4107 Advanced Artificial Intelligence Project is a series of Experiments written in Python. It supports libraries such as Sklearn and Keras.
Abstract — This paper contains an analysis of the accuracy and error rate produced by three standard machine learning algorithms when considering classification as the method of problem-solving for a hand-writing dataset, these algorithms are an Artificial Neural Network (ANN) model, a K-Nearest Neighbour (kNN) model and a combination or hybrid model of the two mentioned models (ANN and kNN). The data that will be analysed in this comparative study it is a hand-writing data called “mnist”, this analysis has the goal to recognize the handwriting elements present in the dataset by making use of the algorithms in cause, the recognition will be done via a classification approach.
The aim or ultimate goal of this comparative study is to evaluate the accuracy and error rate produced by three machine learning algorithms, these algorithms are the Artificial Neural Network (ANN), the K-Nearest Neighbour and the last but not the least a hybrid algorithm created from the initial Artificial Neural Network algorithm and the K-Nearest Neighbour algorithm.
Abstract — In this paper an analysis related to the accuracy and error rate results produced of three machine learning algorithms from Scikit-Learn will be discussed in the context of text classification. The machine learning algorithms that will be discussed are the Recurrent Neural Network (RNN) model acquired from the Python Keras Library, the Multi-Layer Perceptron (MLP) model acquired from Python Scikit-Learn Library and the Support Vector Machines (SVM) model also acquired from the Python Scikit-Learn Library. The type of data that will be analysed in this comparative study it is related to the common text classification problem. The analysis will consist of two different type of problems being investigated, these problems are represented in the form of datasets related to IMDB Movies reviews and sentiment and the identification of SPAM Text Messages in a binary manner. The analysis of the IMDB Movies data has the goal to identify the sentiment present on the review in cause, the sentiment can be either positive or negative, this classification will be represented in a binary manner. The analysis of the SPAM Text Messages data has the goal identify a section of text as either spam or not, this classification or identification of the spam text will be done in a binary manner. The results for the IMDB Movie data represented in the form of accuracy and error rate are approximative 80% accuracy and 20% error rate. The results for the SPAM data, similar to the IMDB Movie data it is represented in the form of accuracy and error rate with values of approximative 90% accuracy and 10% error rate. It is safe to say that the RNN, MLP and the SVM models have performed better on the SPAM Text Messages dataset comparing to the IMDB Movie dataset. This paper also contains a discussion related to Explainable AI topic and the Ethics related to AI projects when it comes to their involvement and impact on human life.
The aim or ultimate goal of this comparative study is to evaluate the accuracy and error rate produced by three machine learning algorithms from the Python Scikit-Learn and Keras Libraries. This paper will present a comparative study related to three machine learning algorithms in the context of text classification, the three machine learning algorithms are from the Python Libraries called Scikit-Learn and Keras. The machine learning algorithms chosen for this comparative study are the Recurrent Neural Network (RNN) model, Multi-Layer Perceptron (MLP) model and the Support Vector Machines (SVM) model. The Recurrent Neural Network (RNN) algorithm has been chosen as Jason Brownlee argues that this type of Neural Network used in the form of Long Short-Term Memory type of Recurrent Neural Network will produce good results on text classification or language translation problems. The Multi-Layer Perceptron algorithm has been chosen as this algorithm it is an Artificial Neural Network (ANN) with multiple layers and it can be defined as a classic algorithm in the Neural Networks area. The Support Vector Machine (SVM) algorithm has been chosen in this study as this algorithm it uses a discrimination approach in order to solve classification problems.
- Python (2.7 required)
- Anaconda (required)
- matplotlib (required)
- sklearn (required)
- pandas (required)
- CX_Freeze (optional)
- Git (optional)
The [CM4107 Advanced Artificial Intelligence Project installation guides] includes instructions for installing the project as part of a local application.
python <path/to/main.py>- Path to entry point file. If unspecified, the current working directory is used.