1 January 2019 in Journal of Computational Social Science
We explore the predictive power of historical news sentiments based on financial market performance to forecast financial news sentiments. We define news sentiments based on stock price returns averaged over one minute right after a news article has been released. If the stock price exhibits positive (negative) return, we classify the news article released just prior to the observed stock return as positive (negative). We use Wikipedia and Gigaword five corpus articles from 2014 and we apply the global vectors for word representation method to this corpus to create word vectors to use as inputs into the deep learning TensorFlow network. We analyze high-frequency (intraday) Thompson Reuters News Archive as well as the high-frequency price tick history of the Dow Jones Industrial Average (DJIA 30) Index individual stocks for the period between 1/1/2003 and 12/30/2013. We apply a combination of deep learning methodologies of recurrent neural network with long short-term memory units to train the Thompson Reuters News Archive Data from 2003 to 2012, and we test the forecasting power of our method on 2013 News Archive data. We find that the forecasting accuracy of our methodology improves when we switch from random selection of positive and negative news to selecting the news with highest positive scores as positive news and news with highest negative scores as negative news to create our training data set.
3 April 2010 in Journal of Economic Interaction and Coordination
22 August 2022 in PLOS ONE
Cryptoassets flow among players as recorded in the ledger of blockchain for all the transactions, comprising a network of players as nodes and flows as edges. The last decade, on the other hand, has witnessed repeating bubbles and crashes of the price of cryptoassets in exchange markets with fiat currencies and other cryptos. We study the relationship between these two important aspects of dynamics, one in the bubble/crash of price and the other in the daily network of crypto, by investigating Bitcoin and XRP. We focus on “regular players” who frequently appear on a weekly basis during a period of time including bubble/crash, and quantify each player’s role with respect to outgoing and incoming flows by defining flow-weighted frequency. During the most significant period of one-year starting from the winter of 2017, we discovered the structure of three groups of players in the diagram of flow-weighted frequency, which is common to Bitcoin and XRP in spite of the different nature of the two cryptos. By examining the identity and business activity of some regular players in the case of Bitcoin, we can observe different roles of them, namely the players balancing surplus and deficit of cryptoassets (Bal-branch), those accumulating the cryptoassets (In-branch), and those reducing it (Out-branch). Using this information, we found that the regime switching among Bal-, In-, Out-branches was presumably brought about by the regular players who are not necessarily dominant and stable in the case of Bitcoin, while such players are simply absent in the case of XRP. We further discuss how one can understand the temporal transitions among the three branches.
13 February 2020 in PLOS ONE
The interactions among macroprices with leads and lags play a significant role in explaining the behavior of an aggregate price index. Thus, to understand inflation and deflation, it is essential to explore the mechanism according to which these macroprices interact with each other. On the basis of a new method, we show that, irrespective of the sources of shocks, a robust flow of changes occurs in domestic prices from upstream to downstream. Moreover, we demonstrate that macroprices change in clusters, and we identify these clusters. Firms are not symmetric. Overall, our analysis suggests that the inertia arising from input/output linkages in production explains the behavior of aggregate prices.
1 September 1999 in Literary and Linguistic Computing
Recent observations in the theory of verse and empirical metrics have suggested that constructing a verse line involves a pattern-matching search through a source text, and that the number of found elements (complete words totalling a specified number of syllables) is given by dividing the total number of words by the mean number of syllables per word in the source text. This paper makes the latter point explicit mathematically, and in the course of this demonstration shows that the word length frequency totals in English prose output are distributed geometrically (previous researchers reported an adjusted Poisson distribution), and that the sequential distribution is random at the global level, with significant non-randomness in the fine structure. Data from a corpus of just under two million words and a syllable-count lexicon of 71,000 word forms is reported, together with some speculations concerning the relationship between the word length frequency distributions in output and in the lexicon. The pattern-matching theory is shown to be internally coherent, and it is observed that some of the analytical techniques described here form a satisfactory test for regular (isometric) lineation in a text.
1 January 2015 in SSRN Electronic Journal
Micro price data shows that individual price settings are not time-invariant as presumed in the existing literature. Furthermore, the analysis of auto-correlati