A comparative study on effect of news sentiment on stock price prediction with deep learning architecture.

The accelerated progress in artificial intelligence encourages sophisticated deep learning methods in predicting stock prices. In the meantime, easy accessibility of the stock market in the palm of one's hand has made its behavior more fuzzy, volatile, and complex than ever. The world is looking at an accurate and reliable model that uses text and numerical data which better represents the market's highly volatile and non-linear behavior in a broader spectrum. A research gap exists in accurately predicting a target stock's closing price utilizing the combined numerical and text data. This study uses long short-term memory (LSTM) and gated recurrent unit (GRU) to predict the stock price using stock features alone and incorporating financial news data in conjunction with stock features. The comparative study carried out under identical conditions dispassionately evaluates the importance of incorporating financial news in stock price prediction. Our experiment concludes that incorporating financial news data produces better prediction accuracy than using the stock fundamental features alone. The performances of the model architecture are compared using the standard assessment metrics -Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Correlation Coefficient (R). Furthermore, statistical tests are conducted to further verify the models' robustness and reliability.

Structure motif-centric learning framework for inorganic crystalline systems.

Incorporation of physical principles in a machine learning (ML) architecture is a fundamental step toward the continued development of artificial intelligence for inorganic materials. As inspired by the Pauling's rule, we propose that structure motifs in inorganic crystals can serve as a central input to a machine learning framework. We demonstrated that the presence of structure motifs and their connections in a large set of crystalline compounds can be converted into unique vector representations using an unsupervised learning algorithm. To demonstrate the use of structure motif information, a motif-centric learning framework is created by combining motif information with the atom-based graph neural networks to form an atom-motif dual graph network (AMDNet), which is more accurate in predicting the electronic structures of metal oxides such as bandgaps. The work illustrates the route toward fundamental design of graph neural network learning architecture for complex materials by incorporating beyond-atom physical principles.

Electrocatalytic CO2 reduction on earth abundant 2D Mo2C and Ti3C2 MXenes.

Mo2C and Ti3C2 MXenes were investigated as earth-abundant electrocatalyts for the CO2 reduction reaction (CO2RR). Mo2C and Ti3C2 exhibited faradaic efficiencies of 90% (250 mV overpotential) and 65% (650 mV overpotential), respectively, for the reduction of CO2 to CO in acetonitrile using an ionic liquid electrolyte. The use of ionic liquid 1-ethyl-2-methylimidazolium tetrafluoroborate as an electrolyte in organic solvent suppressed the competing hydrogen evolution reaction. Density functional theory (DFT) calculations suggested that the catalytic active sites are oxygen vacancy sites on both MXene surfaces. Also, a spontaneous dissociation of adsorbed COOH species to a water molecule and adsorbed CO on Mo2C promote the CO2RR.