Machine learning for financial transaction classification across companies using character-level word embeddings of text fields

An important initial step in accounting is mapping financial transfers to the corresponding accounts. We devised machine-learning-based systems that automate this process. They use word embeddings with character-level features to process transaction texts. When considering 473 companies independently, our approach achieved an average top-1 accuracy of 80.50%, outperforming baselines that exclude the transaction texts or rely on a lexical bag-of-words text representation. We extended the approach to generalizes across companies and even across different corporate sectors. After standardization of the account structures and careful feature engineering, a single classifier trained on 44 companies from 28 sectors achieved a test accuracy of more than 80%. When trained on 43 companies and tested on the remaining one, the system achieved an average performance of 64.62%. This rate increased to nearly 70% when considering only the largest sector.     

Media-Based Sentiment Indices as an Alternative Measure of Consumer Confidence

In this paper, we consider the feasibility of constructing online sentiment indices, using large amounts of media data, as an alternative to the conventional survey method used to create the consumer confidence index in South Africa. A clustering framework is adopted to provide an indication of possible candidate sentiment indices constructed from a combination of different text sources and dictionaries that best mimic the traditional survey-based consumer confidence index from the South African Bureau for Economic Research (BER). The results conclude that it is possible to create an index using sentiment analysis using online editorial data that does resemble the BER’s consumer confidence index. The different media-based sentiment indices (MSI) show a significant level of correlation and co-movement with the BER’s CCI. Impulse responses and cross-correlation functions indicate that the MSI could potentially lead the survey-based method up to two quarters. Furthermore, Granger-causality tests show that the media-based indices are good predictors of future consumer confidence index values. The results provide motivation for further study on the use of sentiment-based techniques and online media data sources to track consumer confidence within an emerging market such as South Africa.     

Psychometric validation of anti-clot treatment scale and treatment satisfaction questionnaire for medication version II in Japanese patients with atrial fibrillation

Aims: The Anti-Clot Treatment Scale (ACTS) and Treatment Satisfaction Questionnaire for Medication version II (TSQM-II) are validated treatment satisfaction patient-reported outcome (PRO) instruments. The ACTS includes two domains: Burdens and Benefits; the TSQM-II includes four: Effectiveness, Side Effects, Convenience, and Global Satisfaction. Japanese-language versions of the ACTS and TSQM-II have been developed and linguistically validated. This study aimed to assess their psychometric properties in Japanese patients with atrial fibrillation (AF).

Materials and methods: ACTS and TSQM-II data from 534 patients with AF were collected in a Japanese post-marketing surveillance study of a direct oral-anticoagulant, rivaroxaban. Four key psychometric properties, in line with best practice guidelines from the US Food and Drug Administration, were examined using traditional psychometric methods: acceptability, scaling assumptions, reliability (i.e. internal consistency reliability, test-retest reliability), and construct validity (i.e. convergent validity and known groups).

Results: ACTS Burdens and Benefits and TSQM-II Effectiveness, Convenience, and Global Satisfaction scales were found to be acceptable (e.g. item-level missing data at baseline <4%), with all scales having good internal consistency (Cronbach’s alpha > 0.80). test-retest reproducibility intraclass correlation coefficients for the ACTS Burdens and Benefits were 0.59 and 0.65, respectively, and between 0.54–0.61 for the TSQM-II scales. Known-groups validity for the ACTS and TSQM-II was supported by differences in scale scores by positive and negative impact (p?<?0.05). Correlations between the ACTS and TSQM-II (convergent validity) were lower than expected (range r?=?0.09–0.48), but in line with the original ACTS development study.

Limitations: Evaluation of test-retest reproducibility was limited by assessment period, which was longer (3 months) than recommended guidelines (usually up to 2 weeks).

Conclusions: Overall, Japanese versions of ACTS and TSQM-II scales satisfied internal consistency reliability and traditional validity criteria. Our study supports the ACTS and TSQM-II as appropriate PRO instruments to measure satisfaction with anticoagulant treatment in Japanese patients with AF.

Trial registration: NCT01598051, clinicaltrials.gov; registered April 20, 2012.     

Predicting Household Poverty: A Methodological Note with a Kenyan Example

Household surveys provide data that is used for identifying and measuring the poverty status of households and individuals. However, carrying out such surveys is expensive, especially in poor developing countries. Thus it is important to make maximum use of the available survey data in developing countries, especially in sub‐Saharan Africa, where such data are expensive to collect and analyse. This paper develops a simple method for using poverty indices derived from survey data for a given year, to predict poverty rates for subsequent periods without having to conduct a new household survey. We illustrate the workings of the method with data from Kenyan household surveys for 1994 and 1997.     

The Local Whittle Estimator of Long-Memory Stochastic Volatility

We propose a new semiparametric estimator of the degree of persistencein volatility for long memory stochastic volatility (LMSV) models.The estimator uses the periodogram of the log squared returnsin a local Whittle criterion which explicitly accounts for thenoise term in the LMSV model. Finite-sample and asymptotic standarderrors for the estimator are provided. An extensive simulationstudy reveals that the local Whittle estimator is much lessbiased and that the finite-sample standard errors yield moreaccurate confidence intervals than the widely-used GPH estimator.The estimator is also found to be robust against possible leverageeffects. In an empirical analysis of the daily Deutsche Mark/USDollar exchange rate, the new estimator indicates stronger persistencein volatility than the GPH estimator, provided that a largenumber of frequencies is used.     

An Examination of the Interaction Between Modelling and Its Relationship with Construction Kits: Lessons from the Past and for the Future

Construction kits have played a significant part in nurturing the growth and development of the minds and manipulation-based skills of children (and adults) in formal and non-formal education settings. These kits have origins rooted in the representation of the built world and now have a diversity of form and function, including technical versions with moving parts. This article examines some of the historically based ideas that lie behind the role that kits may have in terms of physical modelling. The article traces the transformation of kits from simple bricks and blocks into more complicated pieces for space-filling and achieving the transfer of forces and motion. This transformation occurred during the Victorian era when the influence of the ideas of educationalists and, perhaps rather strangely, exponents of early aviation, played major roles in determining the diversity of forms of construction kit and modelled possibilities that are now to be seen in classrooms and homes. The article ends with a review of some of the lessons from the past and present that will need to be considered in relation to concrete modelling in schools for the future. This revised version was published online in July 2006 with corrections to the Cover Date.     

FDI facilitated by agglomeration economies: evidence from manufacturing and services joint ventures in China

The open door policy of China’s economic reform since the 1980s has attracted heavy foreign direct investment (FDI) flows into China and especially to Guangdong (particularly the Pearl River Delta region, PRD) and induced significant economic growth during the past two decades. While there exist various classical theories of FDI in attempting to identify the determinants of FDI inflow and to explain the behavior of FDI flows, limited attention has been given from the perspective of agglomeration effects generated by a core-periphery (CP) relation.This paper intends to study the impacts of agglomerations on FDI inflows in the context of Krugman’s CP relation (1991) by investigating (1) the formation of a CP relation via gravity model analysis; (2) whether different types of industry FDI flows will respond differently in the CP-system, given agglomeration effects; and (3) whether FDI origin and firm scale matter in affecting FDI flows.A database consisting of a population frame of 37,742 firm-level manufacturing and services joint ventures investing in Guangdong in 1998 was used. Empirical results show that the agglomerations of the CP relation have affected FDI flow patterns. While both manufacturing and services FDI and sources of investment responded differently to the impacts, smaller firms were found more responsive to the CP-agglomeration settings regardless of FDI by industry type and by source. The significance and implications of the CP-system to further facilitate FDI in the region are discussed.