دانلود کتاب مبانی و پیشرفت‌های یادگیری ماشین در آمار رسمی

A book on machine learning in official statistics takes on an ambitious goal, as it attempts to combine two subject areas that are complex in themselves. It could therefore be tempting to confine the discussion to the methodological and technical aspects. This is not the option taken in this publication. Rather, it seeks to take a broad approach, addressing the foundations and the advances, the origins and the future. This is convenient as it allows the authors to understand the integration of machine learning as an evolutionary progress in official statistics, which in turn makes it possible to build on the structures that have emerged from earlier stages and on the experience gained from them. Over the past two centuries, official statistics have contributed to the construction of states and have gone through various ups and downs, good and difficult times, in the tides of politics. Developments in statistics usually progress continuously and slowly. However, there are also times when abrupt changes occur, namely, when several of the driving forces are acting simultaneously, through new data, methods, technologies; or through new societal information needs, crises, framework conditions; or through an interaction between such driving forces. The last time statistics industry experienced a major development spurt of this kind was a quarter of a century ago, after the fall of the Berlin Wall and the end of the Cold War, when in the same period personal computers and the internet began to revolutionise the work in statistics factories and in communication with social users and groups. The term ‘statistics industry’ is used here because it allows us to refer to similarities in developments such as the switch to a more decentralised organisation of data processing, which on the bright side led to improved efficiency and acceleration, but on the dark side posed risks to the quality of statistics and challenging adjustments to the structure of personnel in the offices. The origins of a modern understanding of official statistics, namely as an infrastructure for public discourse and the common good based on trustworthy information, go back to this time of upheaval, which was both technical and social. Since then, we have become accustomed to defining the quality of statistics from the end, as ‘fitness for purpose’. Modern management of quality in a comprehensive ‘total’ sense, as well as ethical guidelines, has found their way into international standards and codes, such as the UN Fundamental Principles of Official Statistics, the Declaration of Professional Ethics of the International Statistical Institute and the European Statistics Code of Practice. Organisational structures and processes, the business model and governance for the statistics industry have changed fundamentally, and international cooperation has become even more important. In the flow of developments, we are once again passing through a rapid section with significant currents and shallows, which require fast and attentive manoeuvring. The focus is on the interrelated elements in the technology cluster, which includes machine learning as well as cybersecurity or social media, for example. The socio-political context is similarly important in the form of an erosion of social cohesion, changes in lifestyle and, in general, changes in social institutions caused by information flows, new technologies, echoed in turn by political governance through information. This volume addresses a wide range of issues related to the digital transformation, with a view to the possible improvement of efficiency, timeliness and other aspects of quality while maintaining and preserving the core brand of official statistics. All stages in the value chain of statistics are involved, from design and production to communication and application by users. Developing statistical methodology is translated and deployed in practice, where experts in various fields need to be familiarised with it. Both in terms of data collection and communication, new technologies are opening up possibilities that are being explored. Finally, governance and ethics issues are contained in this book, which provide a framework for responsible, secure and fair handling of the new technologies and data. The statistical community as a whole can best address these issues by working together as quickly and on as international a scale as possible. Technology should be seen as a driver, but not the only one. Equally important are the users and their information needs, as well as the experts who, as producers of statistics, are responsible for ensuring high quality. ‘Zukunft braucht Herkunft’1 (‘Future needs origin’) is taken from an essay by Odo Marquard, a German philosopher. Technologies and crises present themselves in a new configuration, with new challenges, and for this, old recipes are not enough. Without knowledge of the past, without lessons learned from experience, however, one loses one’s compass in the present. Striking a balance between openness to the future and orientation to the existing, driving rapid evolutionary development without shaking the supporting pillars that need to be preserved—that is the challenge and art in this epoch of transformation.