Son yıllarda internet bağlantısı imkanlarının gelişmesiyle, internet kanalı üzerinden yapılan perakende ticaretin büyüklüğü hızla artmıştır. (Euromonitor, 2018) raporu Türkiye’de elektronik ortamda yapılan perakende ticaretin yıllık... more
Son yıllarda internet bağlantısı imkanlarının gelişmesiyle, internet kanalı üzerinden yapılan perakende ticaretin büyüklüğü hızla artmıştır. (Euromonitor, 2018) raporu Türkiye’de elektronik ortamda yapılan perakende ticaretin yıllık ortalama %33 oranında büyüyerek 2022 yılında 75.7 Milyar TL’ye çıkacağını öngörmektedir. Sektörün büyümesinin tahmin edilmesinde kullanılabilecek çok çeşitli makro ekonomik, demografik, finansal öz nitelikler olmasına rağmen tüketici eğilimlerinin modellere yansıtılması önemli bir sorundur. Bu çalışmada, (Kamakura & Du,2012)’nun dinamik faktör analizi yaklaşımı tabanlı bir metot, Türkiye’de farklı alt-sektörlerde faaliyet gösteren çevrim içi perakende markalarının 2014 – 2017 yılları arasındaki haftalık, arama eğilimleri verileri üzerine uygulanmış ve Türkiye’de internet arama eğilimleri açısından sektörün segment yapısı tahmin edilmiştir. İkinci aşamada ise ülkedeki önde gelen çevrim içi perakende markalarından birinin satış rakamlarının tahmini için bir ARIMA modeli kurulmuştur. Takiben, firmanın internet cirosunu öngörmek için kurulan modele, kendi arama trendleri ve faktör analizinden elde edilen arama trendi verisi eklenerek elde edilen, öngörü başarıları karşılaştırılmıştır. Sonuçlar, internet arama eğilimlerinde saklı tüketici eğilimlerinin, ciro öngörüsü üzerinde sınırlı da olsa bir toparlanmaya işaret ettiğini göstermektedir.
the relevant annual cost for agricultural productivity loss is estimated to be around 295 million euros. Under climate changes, soil erosion due to rainfall is dramatically increasing, for the most part because of an increasing of the... more
the relevant annual cost for agricultural productivity loss is estimated to be around 295 million euros. Under climate changes, soil erosion due to rainfall is dramatically increasing, for the most part because of an increasing of the frequency of extreme, localised events. Here, we present the MSCA-Horizon2020 project, focused on understanding and quantifying extreme rainfall effects on soil erosion, by means of ground-based weather-radar observations and hydrological modelling at regional scale (namely in Tuscany, central Italy). In critical hydrological phenomena, such as intense surface runoff, flooding and soil erosion, the spatiotemporal extent is crucial in the development of the processes. This feature significantly affects the impact and the evolution of critical phenomena, especially during extreme events. Therefore, an approach directed to refine as much as possible the knowledge of these dynamics is recommended both at the monitoring and the modelling level. Using an approach based on statistical analyses of rainfall data from ground-based radar and modelling, this project aims to: 1) Quantify on historical data the spatiotemporal distribution of extreme rainfalls / runoff and soil erosion over the last years, 2) Build a platform to model runoff and soil erosion during extreme events in real-time, 3) Simulate in real-time runoff and soil erosion behaviours related to extreme rainfalls, integrating the current regional-warning-system for the extreme weather events.
This paper focus on measuring the performance of algortimic trading in now-casting of stock returns using machine learning technics. For this task, (i) nine commonly used trend indicators to capture the behavior of the stock and a binary... more
This paper focus on measuring the performance of algortimic trading in now-casting of stock returns using machine learning technics. For this task, (i) nine commonly used trend indicators to capture the behavior of the stock and a binary variable to signal positive/negative returs are used as predictors and target variable, respectively; (ii) the standart machine learning process (splitting data, choosing the best performing algorithm among the alternatives, and testing this algorithm for new data) is applied to ASELSAN (a Turkish defense industry company) stock traded in BIST-100. The main findings are: (I) the decission tree algoritm performs better than K-nearest Neighbours, Logistic Regression, Bernoili Naïve Bayes alternatives; (ii) the now-casting model allowed to realize an 18% of yield over the test period; (iii) the model’s performance metrics (accuracy, precision, recall, f1 scores and the ROC-AUC curve) that are commonly used for classification models in machine learning takes values just in the acceptance boundary.
The internet has become one of the most effective means for travellers to seek information on destinations. This paper uses support vector regressions (SVRs) and Google search data to test whether observing internet habits can provide... more
The internet has become one of the most effective means for travellers to seek information on destinations. This paper uses support vector regressions (SVRs) and Google search data to test whether observing internet habits can provide insights on trends in tourist arrivals to Barbados. We find evidence that Google Trends data may be used to (informally) pick up on changing patterns and trends in tourist arrivals from the UK and Canada. But while graphical analysis show some similar patterns in the Google data and reported data on US arrivals, we find no evidence to suggest that Google data provides adds any significant information to what can be “learned” from an autoregressive-SVR.
Age of Big Data and internet has brought variety of opportunities for social researchers on identifying on-going social trends instantly. As internet user base grew exponentially, major internet content search companies have begun to... more
Age of Big Data and internet has brought variety of opportunities for social researchers on identifying on-going social trends instantly. As internet user base grew exponentially, major internet content search companies have begun to offer data mining products which could extract attitude of on-going trends and identify new trends on web as well. Since 2009, as a pioneer on these web analytics solutions Google has launched Google Trends service, which enables to researchers to examine change of trend on specific keywords. We use weekly Google Trends Index of 'General Purpose Loan' (GT) and total out-standing volume of Turkish banking system from the data period of first week of March 2011 to second week of September 2014. In this paper we test whether the Google Analytics search index series can be used as a consistent forecaster of national general purpose loan (GPL] demand in Turkey. We show how to use search engine data to forecast Turkish GPL demand. The results show tha...
In the present study, the effectiveness of nowcasting convective activities using a microwave radiometer has been examined for Kolkata (22.65° N, 88.45° E), a tropical location. It has been found that the standard deviation of brightness... more
In the present study, the effectiveness of nowcasting convective activities using a microwave radiometer has been examined for Kolkata (22.65° N, 88.45° E), a tropical location. It has been found that the standard deviation of brightness temperature (BT) at 22 GHz and instability indices like Lifting Index (LI), K Index (KI) and Humidity Index (HI) has shown definite changes before convective events. It is also seen that combination of standard deviation of BT at 22 GHz and LI can be most effective in predicting convection. A nowcasting algorithm is prepared using 18 isolated convective events of 2011 and in all cases, a marked variation of these parameters has been seen an hour before the event. Accordingly, a prediction model is developed and tested on convective events of 2012 and 2013. It is seen that the model gives reasonable success in predicting convective rain about 7075 min in advance with a prediction efficiency of 80%.
Age of Big Data and internet has brought variety of opportunities for social researchers on identifying on-going social trends instantly. As internet user base grew exponentially, major internet content search companies have begun to... more
Age of Big Data and internet has brought variety of opportunities for social researchers on identifying on-going social trends instantly. As internet user base grew exponentially, major internet content search companies have begun to offer data mining products which could extract attitude of on-going trends and identify new trends on web as well. Since 2009, as a pioneer on these web analytics solutions Google has launched Google Trends service, which enables to researchers to examine change of trend on specific keywords. We use weekly Google Trends Index of “General Purpose Loan” (GT) and total out-standing volume of Turkish banking system from the data period of first week of March 2011 to second week of September 2014. In this paper we test whether the Google Analytics search index series can be used as a consistent forecaster of national general purpose loan (GPL) demand in Turkey. We show how to use search engine data to forecast Turkish GPL demand. The results show that Google search query data is successful at nowcasting GPL demand.
In this paper, a novel Radar-based technique to forecast probability of rainfall within the next 4 hours is presented. Time series data of Doppler Weather RADARs located at different points in the Philippines allow us to observe and... more
In this paper, a novel Radar-based technique to forecast probability of rainfall within the next 4 hours is presented. Time series data of Doppler Weather RADARs located at different points in the Philippines allow us to observe and forecast the chance of rain at a more localized level. Trajectories derived from the trail of past observations are used to generate the forecasted location of rainfall. Based on their projected locations, a percent chance of rain (PCOR) per city is calculated. From this technique, automatically obtained forecasts for rain events are accurate with an average accuracy of 82.68%, and with an average success ratio of 57.98% peaking at 76% at the first hour for forecasted rain with actual rain events.
The World Meteorological Organization (WMO) World Weather Research Programme’s (WWRP) Forecast and Research in the Olympic Sochi Testbed program (FROST-2014) was aimed at the advancement and demonstration of state-of-the-art nowcasting... more
The World Meteorological Organization (WMO) World Weather Research Programme’s (WWRP) Forecast and Research in the Olympic Sochi Testbed program (FROST-2014) was aimed at the advancement and demonstration of state-of-the-art nowcasting and short-range forecasting systems for winter conditions in mountainous terrain. The project field campaign was held during the 2014 XXII Olympic and XI Paralympic Winter Games and preceding test events in Sochi, Russia. An enhanced network of in situ and remote sensing observations supported weather predictions and their verification. Six nowcasting systems (model based, radar tracking, and combined nowcasting systems), nine deterministic mesoscale numerical weather prediction models (with grid spacings down to 250 m), and six ensemble prediction systems (including two with explicitly simulated deep convection) participated in FROST-2014. The project provided forecast input for the meteorological support of the Sochi Olympic Games. The FROST-2014 archive of winter weather observations and forecasts is a valuable information resource for mesoscale predictability studies as well as for the development and validation of nowcasting and forecasting systems in complex terrain. The resulting innovative technologies, exchange of experience, and professional developments contributed to the success of the Olympics and left a post-Olympic legacy.
We apply a soft computing method to generate country-specific economic sentiment indicators that provide estimates of year-on-year GDP growth rates for 19 European economies. First, genetic programming is used to evolve business and... more
We apply a soft computing method to generate country-specific economic sentiment indicators that provide estimates of year-on-year GDP growth rates for 19 European economies. First, genetic programming is used to evolve business and consumer economic expectations to derive sentiment indicators for each country. To assess the performance of the proposed indicators, we first design a nowcasting experiment in which we recursively generate estimates of GDP at the end of each quarter, using the latest business and consumer survey data available. Second, we design a forecasting exercise in which we iteratively re-compute the sentiment indicators in each out-of-sample period. When evaluating the accuracy of the predictions obtained for different forecast horizons, we find that the evolved sentiment indicators outperform the time-series models used as a benchmark. These results show the potential of the proposed approach for prediction purposes.
This paper discusses the use of Random Forest (RF), a popular Machine Learning (ML) algorithm, to perform spatially explicit nowcasting of cloud-to-ground lightning occurrence. An application to the Italian territory and the surrounding... more
This paper discusses the use of Random Forest (RF), a popular Machine Learning (ML) algorithm, to perform spatially explicit nowcasting of cloud-to-ground lightning occurrence. An application to the Italian territory and the surrounding seas is presented. Specifically, 1-hour ahead lightning occurrences over the months of August, September and October from 2017 to 2019 have been modelled using a dataset including geo-environmental features. Results obtained with three different spatial resolutions have been compared, for nowcasting both positive and negative strokes. The features’ importance resulting from the best RF models showed how datadriven models are able to identify the relationships between meteorological variables, in agreement with previous physically based knowledge of the phenomenon. The encouraging results obtained in terms of forecasting accuracy support the idea to use ML-based algorithms in early warning procedures for disaster risk management.
Since China’s enactment of the Reform and Opening-Up policy in 1978, China has become one of the world’s fastest growing economies, with an annual GDP growth rate exceeding 10 % between 1978 and 2008. But in 2015, Chinese GDP grew at 7 %,... more
Since China’s enactment of the Reform and Opening-Up policy in 1978, China has become one of the world’s fastest growing economies, with an annual GDP growth rate exceeding 10 % between 1978 and 2008. But in 2015, Chinese GDP grew at 7 %, the lowest rate in five years. Many corporations complain that the borrowing cost of capital is too high. This paper constructs Chinese Divisia monetary aggregates M1 and M2, and, for the first time, constructs the broader Chinese monetary aggregates, M3 and M4. Those broader aggregates have never before been constructed for China, either as simple-sum or Divisia. The results shed light on the current Chinese monetary situation and the increased borrowing cost of money. GDP data are published only quarterly and with a substantial lag, while many monetary and financial decisions are made at a higher frequency. GDP nowcasting can evaluate the current month’s GDP growth rate, given the available economic data up to the point at which the nowcasting is...
This paper evaluates five models of Nowcasting that forecast Mexico's quarterly GDP: a Dynamic Factor Model (MFD), two Bridge Equation Models (BE) and two Principal Components Models (PCA). The results indicate that the BE forecasts... more
This paper evaluates five models of Nowcasting that forecast Mexico's quarterly GDP: a Dynamic Factor Model (MFD), two Bridge Equation Models (BE) and two Principal Components Models (PCA). The results indicate that the BE forecasts average is statistically better than the rest of the models considered, according to the accuracy test of Diebold-Mariano (1995). In addition, using real-time information, it is found that the BE average is more accurate than the median of the forecasts provided by the analysts surveyed by Bloomberg and the median of the experts who answer Banco de México’s Survey of Professional Forecasters.
I evaluate five nowcasting models that I used to forecast Mexico's quarterly GDP in the short run: a dynamic factor model (DFM), two bridge equation (BE) models and two models based on principal components analysis (PCA). The results... more
I evaluate five nowcasting models that I used to forecast Mexico's quarterly GDP in the short run: a dynamic factor model (DFM), two bridge equation (BE) models and two models based on principal components analysis (PCA). The results indicate that the average of the two BE forecasts is statistically better than the rest of the models under consideration , according to the Diebold-Mariano accuracy test. Using real-time information, I show that the average of the BE models is also more accurate than the median of the forecasts provided by the analysts surveyed by Bloomberg, the median of the experts who answer Banco de México's Survey of Professional Forecasters and the rapid GDP estimate released by INEGI.
Wind power forecasting can enhance the value of wind energy by improving the reliability of integrating this variable resource and improving the economic feasibility. The National Center for Atmospheric Research (NCAR) has collaborated... more
Wind power forecasting can enhance the value of wind energy by improving the reliability of integrating this variable resource and improving the economic feasibility. The National Center for Atmospheric Research (NCAR) has collaborated with Xcel Energy to develop a multifaceted wind power prediction system. Both the day-ahead forecast that is used in trading and the short-term forecast are critical to economic decision making. This wind power forecasting system includes high resolution and ensemble modeling capabilities, data assimilation, now-casting, and statistical postprocessing technologies. The system utilizes publicly available model data and observations as well as wind forecasts produced from an NCAR-developed deterministic mesoscale wind forecast model with real-time four-dimensional data assimilation and a 30-member model ensemble system, which is calibrated using an Analogue Ensemble Kalman Filter and Quantile Regression. The model forecast data are combined using NCAR&#...
This paper describes a new multi-sensor approach for convective rain cell continuous monitoring based on rainfall derived from Passive Microwave (PM) remote sensing from the Low Earth Orbit (LEO) satellite coupled with Infrared (IR)... more
This paper describes a new multi-sensor approach for convective rain cell continuous monitoring based on rainfall derived from Passive Microwave (PM) remote sensing from the Low Earth Orbit (LEO) satellite coupled with Infrared (IR) remote sensing Brightness Temperature (TB) from the Geosynchronous (GEO) orbit satellite. The proposed technique, which we call Precipitation Evolving Technique (PET), propagates forward in time and space the last available rain-rate (RR) maps derived from Advanced Microwave Sounding Units (AMSU) and Microwave Humidity Sounder (MHS) observations by using IR TB maps of water vapor (6.2 μm) and thermal-IR (10.8 μm) channels from a Spinning Enhanced Visible and Infrared Imager (SEVIRI) radiometer. PET is based on two different modules, the first for morphing and tracking rain cells and the second for dynamic calibration IR-RR. The Morphing module uses two consecutive IR data to identify the motion vector to be applied to the rain field so as to propagate it in time and space, whilst the Calibration module computes the dynamic relationship between IR and RR in order to take into account genesis, extinction or size variation of rain cells. Finally, a combination of the Morphing and Calibration output provides a rainfall map at IR space and time scale, and the whole procedure is reiterated by using the last RR map output until a new MW-based rainfall is available. The PET results have been analyzed with respect to two different PM-RR retrieval algorithms for seven case studies referring to different rainfall convective events. The qualitative, dichotomous and continuous assessments show an overall ability of this technique to propagate rain field at least for 2–3 h propagation time.
At Paranal Observatory, the least predictable parameter affecting the short-term scheduling of astronomical observations is the optical turbulence, especially the seeing, coherence time and ground layer fraction. These are critical... more
At Paranal Observatory, the least predictable parameter affecting the short-term scheduling of astronomical observations is the optical turbulence, especially the seeing, coherence time and ground layer fraction. These are critical variables driving the performance of the instruments of the Very Large Telescope (VLT), especially those fed with adaptive optics systems. Currently, the night astronomer does not have a predictive tool to support him/her in decision-making at night. As most service-mode observations at the VLT last less than two hours, it is critical to be able to predict what will happen in this time frame, to avoid time losses due to sudden changes in the turbulence conditions, and also to enable more aggressive scheduling. We therefore investigate here the possibility to forecast the turbulence conditions over the next two hours. We call this "turbulence nowcasting", analogously with weather nowcasting, a term already used in meteorology coming from the cont...
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for... more
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for the spatio-temporal prediction of complex problems. In this research, we propose an LSTM convolutional neural network (CNN-LSTM) architecture for immediate prediction of various short-term precipitation events using satellite data. The CNN-LSTM is trained with NASA Global Precipitation Measurement (GPM) precipitation data sets, each at 30-min intervals. The trained neural network model is used to predict the sixteenth precipitation data of the corresponding fifteen precipitation sequence and up to a time interval of 180 min. The results show that the increase in the number of layers, as well as in the amount of data in the training data set, improves the quality of the forecast.
In this study, we investigate the ability of several convective initiation predictors based on satellite infrared observations to distinguish convective weak precipitation events from those leading to intense rainfall. The two types of... more
In this study, we investigate the ability of several convective initiation predictors based on satellite infrared observations to distinguish convective weak precipitation events from those leading to intense rainfall. The two types of precipitation are identified according to hourly rainfall, respectively less than 10 mm and greater than 30 mm. The analysis is conducted on a representative dataset containing 92 severe and weak precipitation events collected over the Italian peninsula in the period 2016–2019 over June-September. The events are selected to be short-lived (i.e., less than 12 h) and localized (i.e., less than 50×50km2). Italian National Radar Network products, namely the Vertical Maximum Intensity (VMI) and the Surface Rain Total (SRT) variables (from Dewetra Platform by CIMA, Italian Civil Protection Department), are used as indicators of convection (i.e., VMI greater than 35 dBZ echo intensity) and cumulated rainfall, respectively. The considered predictors are linea...
This paper describes a new multi-sensor approach for convective rain cell continuous monitoring based on rainfall derived from Passive Microwave (PM) remote sensing from the Low Earth Orbit (LEO) satellite coupled with Infrared (IR)... more
This paper describes a new multi-sensor approach for convective rain cell continuous monitoring based on rainfall derived from Passive Microwave (PM) remote sensing from the Low Earth Orbit (LEO) satellite coupled with Infrared (IR) remote sensing Brightness Temperature (TB) from the Geosynchronous (GEO) orbit satellite. The proposed technique, which we call Precipitation Evolving Technique (PET), propagates forward in time and space the last available rain-rate (RR) maps derived from Advanced Microwave Sounding Units (AMSU) and Microwave Humidity Sounder (MHS) observations by using IR TB maps of water vapor (6.2 μm) and thermal-IR (10.8 μm) channels from a Spinning Enhanced Visible and Infrared Imager (SEVIRI) radiometer. PET is based on two different modules, the first for morphing and tracking rain cells and the second for dynamic calibration IR-RR. The Morphing module uses two consecutive IR data to identify the motion vector to be applied to the rain field so as to propagate it...
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for... more
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for the spatio-temporal prediction of complex problems. In this research, we propose an LSTM convolutional neural network (CNN-LSTM) architecture for immediate prediction of various short-term precipitation events using satellite data. The CNN-LSTM is trained with NASA Global Precipitation Measurement (GPM) precipitation data sets, each at 30-min intervals. The trained neural network model is used to predict the sixteenth precipitation data of the corresponding fifteen precipitation sequence and up to a time interval of 180 min. The results show that the increase in the number of layers, as well as in the amount of data in the training data set, improves the quality of the forecast.
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for... more
The short-term prediction of precipitation is a difficult spatio-temporal task due to the non-uniform characterization of meteorological structures over time. Currently, neural networks such as convolutional LSTM have shown ability for the spatio-temporal prediction of complex problems. In this research, we propose an LSTM convolutional neural network (CNN-LSTM) architecture for immediate prediction of various short-term precipitation events using satellite data. The CNN-LSTM is trained with NASA Global Precipitation Measurement (GPM) precipitation data sets, each at 30-min intervals. The trained neural network model is used to predict the sixteenth precipitation data of the corresponding fifteen precipitation sequence and up to a time interval of 180 min. The results show that the increase in the number of layers, as well as in the amount of data in the training data set, improves the quality of the forecast.
Nowcasting of heavy rain events using microwave radiometer has been carried out at Kolkata (22.65N, 88.45E), a tropical location. Microwave radiometer can produce the temperature and humidity profiles of the atmosphere with fairly good... more
Nowcasting of heavy rain events using microwave radiometer has been carried out at Kolkata (22.65N,
88.45E), a tropical location. Microwave radiometer can produce the temperature and humidity profiles
of the atmosphere with fairly good accuracy. Definite changes are observed in temperature and humidity
profiles before and at the onset of heavy rain events. Concurrent changes in the brightness temperatures
(BT) at 22 GHz and 58 GHz are found to be suitable to nowcast rain. The time derivatives of brightness
temperatures at 22 GHz and 58 GHz are used as inputs to the proposed nowcasting model. In addition,
the standard deviation of the product of these time derivatives is also considered. The model has been
developed using the data of 2011 and validated for rain events of 2012–2013 showing a prediction effi-
ciency of about 90% with alarm generated about 25 min in advance.