This paper is an empirical study of the capitalization rates for office buildings in the core six wards of Tokyo from 2001 to 2009. The purpose of this study is to estimate the spread of the capitalization rates based on accessibility to public parks by a gravity model. It found the following results: 1) the coefficient of the accessibility to public parks that is five hectares and above is statistically significant; and 2) the spreads of transaction-based cap rates are estimated from 10 bps to 40 bps by a ‘with–without a factor’ analysis. It indicated that urban green space amenity affects on office building prices. The results of the study should provide insights to the DCF analysis in appraisal reports.

Taking into account the need to build a novel model to predict corporate credit ratings from published financial information for assessing the defaulting risk of uncertified companies, a large-scale financial data set was analyzed to reproduce ratings for all certified companies with a modern type of data mining technique, called support vector machine (SVM). The regression function included in the LIBSVM software was applied to multiclass classification for credit ratings. Data of 18,119 records of 11 years between 2000 and 2011 year for 1,213 companies published in “Kaishashikiho” were used. Separate SVM models were constructed for 29 corporate groups using 91 to 120 kinds of effective indices created from 13 to 15 kinds of published indices, respectively. The models were optimized by a cross-validation method, and the prediction power was evaluated by using the recent financial data. It was found that the actual ratings given by the rating agencies were reproduced with an accuracy of 86% for the prediction set from 2008 to 2011 year. This demonstrates that the present model effectively functions as a practical tool because of its excellent explanatory and predictive performance from just published available financial indices as input data.