Prediction of Stock Market Price using Bi-directional Recurrent Neural Network

Published: Nov. 2, 2023

Recently, an accurate prediction of stock market returns is a very challenging task due to volatile and non-linear nature the financial markets. The advent artificial intelligence enhanced processing power has led realisation that preprogrammed techniques are more effective at forecasting values. currently one most researched fastest developing subj ects, predicting its behaviour crucial. primary challenge in this area research remains improving forecast precision. Strong shareholder made possible by price forecasting. Despite this, researchers use these algorithms predict continuing trends based on Stock Technical Indicators (STIs), recent advancements machine learning techniques. This study analyses Yahoo Finance data from STIs over decade prices using Bi- directional Recurrent Neural Network (BI-RNN). Initially, analysed were used as input for autoencoder during dimensionality reduction procedure, which reduced correlation between STIs. These then provided BI-RNN. Next, order prices, BI-RNN's outcome attributes fed into soft max layer. From results, it clearly shows proposed Bi-RNN shown better results terms minimal MAPE value 0.41, MAE 4.42, RMSE 0.10, MSE 212.25 experiments, strategy surpassed traditional methods.

Language: Английский

An Empirical Study on: Time Series Forecasting of Amazon Stock Prices using Neural Networks LSTM and GAN models

Published: Nov. 23, 2023

The OHLCV (Open, High, Low, Close, Volume) data used in this study is to forecast time series and anticipate stock price movement. We investigate a wide variety of models, including traditional statistical approaches cutting-edge deep learning strategies combined with sentiment analysis, feature extraction, hyperparameter tweaking. Instead focusing on absolute prices, our main goal predict swings as has been shown produce more accurate outcomes. start research by obtaining historical Amazon via the Yahoo API, then we go thorough analytical journey. generate features first, design test Fourier Autoregressive Integrated Moving Average (ARIMA) models. switch sophisticated methods, using pre-processed apply Long Short-Term Memory (LSTM) Interestingly, add analysis LSTM study, which expands its scope lets us consider market possible influencing factor. To guarantee stability use careful train-test split technique organize manner. field financial forecasting trading methods will ultimately benefit from insightful information study's findings provide efficacy different modeling techniques their capacity movements.

Language: Английский

Comprehensive study on deep-learning-based online course review analysis

Published: Dec. 22, 2023

Under the impact of pandemic, acceptance toward online education increased. Therefore, we have witnessed increasing requirements to help public determine quality courses. This research is related sentiment analysis feedback from course. During process, utilized 458,280 reviews Coursera, across time 2019 2020. First, prepare for deep learning, were transformed by TF-IDF feature. BiLSTM, Transformer (BERT-based), and LSTM with attention mechanisms tested on dataset. The LSTM+attention model produced a result precision 95.41% F1 score 95.48%. context course analysis, this study indicates effectiveness attention.

Language: Английский