ItemAnnex 19 : predicting population-level socio-economic characteristics using Call Detail Records (CDRs) in Sri Lanka(2018) Surendra, Aparna; Lokanathan, Sriganesh; Fernando, Lasantha; Perera-Gomez, ThavishaNational census information is time-consuming and expensive to collect. This research helps determine whether mobile phone data can provide a reliable, cheap proxy for census data within Sri Lanka, especially where post-conflict regions need more frequent data collection. Study findings suggest that socio-economic levels (SEL) can affect call detail records (CDR) data in a post-conflict, Sri Lankan setting. Analysis demonstrates the potential for telecom data to predict census features. The results correspond to assumptions about the population under study, which includes a high percentage of vulnerable, highly mobile groups displaced due to conflict. ItemAnnex 19 : predictive model for the dengue incidences in Sri Lanka using mobile network big data(2018) Dharmawardana, K.G.S.; Lokuge, K. S.; Dassanayake, P. S. B.; Sirisena, M. L.; Fernando, Lasantha; Perera, Amal Shehan; Lokanathan, SriganeshThe study constructs a usable predictive model for any given Medical Officer of Health (MOH) division, which is the smallest medical administrative district in Sri Lanka, by taking human mobility into account. It includes the importation of dengue into immunologically ’naive’ regions. Derived mobility values for each region of the country are weighted using reported past dengue cases. The study introduces a generalizable methodology to fuse big data sources with traditional data sources, using machine learning techniques. Mobile Network Big Data (MNBD) consists of data categories such as Call Detail Records (CDR), Internet access usage records, and airtime recharge records. ItemFinal technical report : leveraging mobile network big data for developmental policy(2017) LIRNEasiaThe research addresses how big data can provide evidence to better inform public policy and allow for greater use of evidence in the policy making process. In addition to more detailed research in the area of transportation and urban planning (commuting patterns), this research articulates and answers questions in other domains such as health (modeling the spread of diseases) and official statistics (mapping poverty for instance). Guidelines were translated into legal language so that mobile operators can responsibly share data. Traditional survey methods that provide enough detail to accurately assess conditions are costly and can rarely reach a representative portion of the population, especially in poorer areas. ItemLeveraging mobile network big data for developmental policy : final technical report(2018-03)This final report reviews research conducted around the potential uses of big data towards better informed public policy, and to expand the use of evidence-based research in the policy making process. Through this project, LIRNEasia extended its pioneering research based on Mobile Network Big Data (MNBD). New sources of data were leveraged in addition to MNBD, including Closed Circuit Television (CCTV), satellite imagery, electricity data, crowd-sourced data, as well as “small” data from government agencies and others. In addition to the generation of actionable insights, LIRNEasia is developing capacity in local universities and among new researchers. ItemImproving disease outbreak forecasting models for efficient targeting of public health resources(2018-03) Fernando, Lasantha; Perera, Amal; Lokanathan, SriganeshThe forecasting models developed in this work can be utilized to effect better resource mobilisation for combatting dengue. For understanding human mobility in disease propagation, Mobile Network Big Data (MNBD) is a low cost data exhaust that provides rich insight into human mobility patterns, including better spatial and temporal granularity. Research focuses on the development of a human mobility model, using MNBD that can accurately depict aggregate human population movements in Sri Lanka, and from this determine which machine learning technique provides the best disease forecasting model. ItemAnnex 14 : bulk data : policy implications (draft)(2018-03) Samarajiva, Rohan; Perera-Gomez, ThavishaThe term “bulk surveillance” is used to describe the collection and analysis of behavioral big data relevant to maintenance of law and order, broadly defined. Avoidance of detection by law breakers may be perceived as easier in virtual space when agents of the law are at a technological disadvantage. The focus of this paper is on the subset of big data known as transaction-generated data (also described as “data exhaust”) arising from the day-to-day behaviors of persons and the technological devices closely associated with them. What should the principles be with regard to bulk surveillance and uses of personal data? ItemAnnex 15 : analyzing Facebook data to understand regional connectivity(2018-03) Wijeratne, Yudhanjaya; Samarajiva, Rohan; Lokanathan, Sriganesh; Surendra, Aparna; Fernando, LasanthaSocial network connectivity between nations potentially serves as a way of exploring networks of international communities, human movement and economic activity. This connectivity often appears to transcend governmental and political barriers (such as between India and Pakistan, for example). Initial exploratory analysis reveals that social network data does strongly correlate to trade and migration between countries, but not to the distance between countries. Some prospective challenges are also outlined for this study. ItemAnnex 16 : automated traffic monitoring for complex road conditions(2018-03) Opatha, R. K; Peiris, Akila; Gamini, D. D. A.; Edirisuriya, Ananda; Athuraliya, C. D.; Jayasooriya, IsuruRecent advancements in computer vision and machine learning techniques have made traffic monitoring systems highly effective in well structured traffic conditions such as highways. But these systems struggle in handling complex and irregular conditions that exist in developing countries, due to lack of infrastructure and regulation. This research breaks down the problem into different sub-tasks such as vehicle detection, vehicle tracking, and vehicle recognition, then combines each process into one pipeline that can be used for traffic monitoring. Implementing the final pipeline involves improving and aggregating existing techniques. Results demonstrate the potential of these techniques for automated traffic monitoring. ItemAnnex 17 : deep semantic segmentation for built-up area extraction and mapping from satellite imagery(2018-03) Athuraliya, C. D.; Ramasinghe, Sameera; Lokanathan, SriganeshResearch focuses on generating more usable built-up area maps, as traditional methods (such as surveys and census) are infrequent and costly. The work proposes a modified Fully Convolutional Network (FCN) architecture that will improve semantic segmentation operation on satellite imagery for built-up area extraction and urban mapping. This method could bridge the gap between existing extraction techniques and actual land cover/built-up area maps used by practitioners. Applications are potentially to socio-economic classification and urban planning, where building density functions as a proxy measure for socio-economic level, and building distribution for urban area estimates and growth, respectively. ItemAnnex 20 : code of practice for the secondary use of mobile network big data(2018-03)Personal data must be protected against accidental destruction or loss, alteration, and unauthorized disclosure or access. This Code of Practice addresses issues related to the processing and uses of Mobile Network Big Data (MNBD), where data collection practices “will be transparent and will not go beyond /will desist from collecting more data than needed for the efficient operation of the network and the supply of goods and services to the customer.” It covers data storage, as well as accountability of Data Controllers, and compliance with national standards within binding agreements. ItemAnnex 21 : scrutiny of electricity billing and supply data as a probable proxy for economic activities : an analysis of power consumption of Dhaka, Bangladesh (draft)(2018-03) Zaber, Moinul; Bhyiyan, Farhad; Sayeed, Abu; Islam, Samiul; Rakib, Nibras; Ali, AminThis case study attempts to provide a load forecasting model to help ascertain short-term electricity demand at the regional level in Bangladesh. To assist policy makers in determining how regulatory decisions impact behavior, consumer level billing data, and power satiation level, supply data such as load variability and load shedding is analyzed. Cleaning the dataset and dealing with outlier values includes such problems as lack of exact household addresses in Dhaka city. The impact of changes in appliance use due to weather or price hikes is examined in order to predict future energy needs of consumers. ItemUsing mobile network big data for land use classification CPRsouth 2015(LIRNEasia, Colombo, LK, 2015-07) Madhawa, Kaushalya; Lokanathan, Sriganesh; Maldeniya, Danaja; Samarajiva, RohanThe traditional way of generating insights on land use involve surveys and censuses, which are both infrequent as well as costly. This paper explores the potential of leveraging massive amounts of human mobile phone data to understand the spatiotemporal activity of mass populations, and by extension, provide a useful proxy for activity-based classification of land use. Understanding and monitoring land use characteristics is critical for urban planning. The study demonstrates possibilities for use of mobile network big data, and how it can be leveraged to infer three distinct land use characteristics: commercial/ economic, residential, and mixed-use. ItemUnderstanding communities using mobile network big data CPRsouth 2015(LIRNEasia, Colombo, LK, 2015-07) Madhawa, Kaushalya; Lokanathan, Sriganesh; Samarajiva, Rohan; Maldeniya, DanajaUnderstanding the strength and boundaries of human connections can help identify communities amongst a population, and is valuable knowledge for modeling disease spread, information flow, and mobility patterns. Administrative boundaries, formed by history and geography, do not necessarily reflect the actual communities or social interaction patterns within a region. In this study we employ community detection algorithms to a mobile Call Detail Records (CDR) network in Sri Lanka in order to compare natural communities existing in the interaction network against administrative regions of Sri Lanka. Additionally we explore how these communities segment into a further level of sub-communities. ItemBig data at the heart of smart cities(Wijeya Newspapers, 2015-09) Samarajiva, RohanUntil recently, constraints of computer memory, retrieval, and processing limited the use of data to entities who could afford supercomputers. Since hardware and memory have declined in price and improved functionality and open-source software has been developed, big data analytics have been democratized. For example, using smart phone data, Sri Lankan city of Colombo has analysed population nodes, and unlike expensive industry surveys, can pinpoint locales as “leaning commercial” or “leaning residential.” To feed data, infrastructure investments are required. But a city becomes smart only when its functioning improves due to enhanced feedback, and creative responses are made. ItemWhere did you come from? : where did you go?; robust policy relevant evidence from mobile network big data(LIRNEasia, Colombo, LK, 2015-03) Maldeniya, Danaja; Kumarage, Amal; Lokanathan, Sriganesh; Kreindler, Gabriel; Madhawa, KaushalyaThe paper discusses how output from mobility analysis based on mobile network big data (MNBD) can be aligned with the different stages of traditional forecasting frameworks familiar to transport planners and policy makers. Levels of accuracy and detail are estimated, so that mobility insights-based MNBD can be delivered. Recently developed approaches for estimating mobility are compared, and results are validated against data from traditional methods. The limitations of MNBD are presented, and alternatives are proposed to address these limitations in future work. The research aims to extend state of the art data mining to support and transform efficiencies in transportation planning.