Workshop on Situation Recognition by Mining Temporal Information SIREMTI2015

Location: Berlin, Germany

co-located with: 7th International Conference on Mobile Computing, Applications and Services MobiCASE 2015


We are happy to announce the invited talk of Prof. Stephan Sigg, Aalto University Finland, who is focusing on collective activity recognition. His talk on Contact-free situation recognition from temporal data will be held during the workshop sessions.


Furnished with wearables to serve our very needs, we traverse sensor-rich environments of developing smart environments and smart cities. This loose connection of wearables and environmental devices, collectively sense social and environmental information. The talk will discuss potential opportunities enabled by this scenario. In with respect to the advance of sentiment sensing, smart city architectures as well as autonomous intelligent spaces. The technologies discussed are woven around a common sensing interface shared by all these wearable and environmental sensing devices: the RF-interface. Ubiquitously available signals from e.g. FM-radio, WiFi or UMTS can be exploited for the detection of presence, location, crowd-size, activity or gestures. On-site training of these systems will soon be replaced by offline- raytracing techniques and recognition accuracies will further increase with the Channel State Information (CSI) on recent OFDM receivers. Sensing of activities or surroundings has been largely explored in recent years. However, this can only incompletely describe a situation. Emotion, attention and intention is what gives meaning to particular human actions.

                 09:00 - 10:00 Registration, Keynote (if you register short before 9:00, you are free to listen to the conference keynote which starts at 9:00)

                10:00 - 10:15 Coffee break

                10:15 - 11:00 Welcome, Introduction, "Trend Mining for Situation Recognition" Olga Streibel

                11:00 - 11:30 "Thread model based security for wireless networks " Freshta Popalyar
Wireless Mesh Network (WMN) is a technology, which has gained popularity due to its cost effective design, robustness, and reliable service coverage. Despite the advantages, WMNs are considered vulnerable to security breaches. Thereby, it is important to consider security in the early design phase in WMNs. Identifying security threats helps the system designer in developing rational security requirements. In this paper we propose threat modeling as a systematic approach to pinpoint the security threats for WMNs as basis for developing security requirements. We identify assets, value them and categorize possible attacks that target the assets in a layer-wise manner. We further elucidate our threat model by use of Attack Trees to clearly define vulnerabilities in the system during early design phase. We take the example of Schools' WMN in a district of Kabul City in Afghanistan as our scenario. We briefly discuss how to assess the risks that are associated with the specified WMN based on the information that is derived from the threat model.

                11:30 - 12:00 "Managing wireless mesh networks: study of recent fault recovery" Akmal Yaqini
Wireless Mesh Network (WMN) is a technology which has evolved in recent years and fits well in today's technological needs. However, due to the wireless nature of WMNs and their deployment in heterogeneous and large scale areas, wireless links often face significant quality fluctuations and performance degradation or weak connectivity. Therefore, failure detection and recovery plays crucial role in performance of WMN. This paper presents a study report on comparison of recent research and techniques developed for the issue of fault tolerance in WMNs. In this survey we present the existing techniques for fault tolerance in WMNs in categories; node failure approach, communication failure approach, routing schemes, fault tolerance techniques, and autonomous reconfiguration systems. The paper also provides an outline of areas which need further research and studies.

                12:00 - 13:30  Lunch break 

Afternoon session

                13:30 - 14:00 "A spatiotemporal approach for social situation recognition" 
                Christian Meurisch, Tahir Hussain, Artur Gogel, Benedikt Schmidt, Immanuel Schweizer, Max Mühlhäuser
The development of virtual personal assistants requires situation awareness. For this purpose, lightweight approaches for the processing of sensor data to derive situation information from available sensor data (e.g., mobile phone data) are required. In this paper, we propose a spatiotemporal approach to derive situational information about social interactions only based on location and time, using data collected with off-the-shelf smartphones. We examine the approach, using location traces of 163 users collected over four weeks. The proposed spatiotemporal approach shows an average social situation recognition result of 45.8 +/- 23.2% F1-measure across the data set using Random Forest classifiers.
                14:00 - 14:30  "Using interaction signals for job recommendations"  Benjamin Kille, Fabian Abel, Balzs Hidasi, Sahin Albayrak 
 Job recommender systems depend on accurate feedback to improve their suggestions. Implicit feedback arises in terms of clicks, bookmarks and replies. We present results from a member inquiry conducted on a large-scale job portal. We analyse correlations between ratings and implicit signals to detect situations where members liked their suggestions. Results show that replies and bookmarks reflect preferences much better than clicks.

                14:30 - 15:00 "SpaceScan: A Supervised Approach to Role Discovery" Ahmad Hasan, Clemens Klein-Robbenhaar, Adrian Paschke
The aim of role discovery is to identify users who exhibit similar behavior within their communities. In this paper, we present a simplified approach to role discovery that helps domain experts identify roles in their domains starting from simple observations. The approach starts with two normalization steps to give more weight to characteristic features of each user. We apply our approach on log files from a collaborative content management system and present the initial results of our data analysis.

                15:00 - 15:30 Coffee break
                15:30 Closing


How can we harvest the information from our environment, including the web, and create solutions to different situation recognition problems,  especially small and big emergency situations of any kind? What if our environment would be equipped with intelligent, web-based assistive devices which, aware of any emergency, would support us and enable an easier and safer living in everyday situations?

Indeed, our cars are already equipped with intelligent driver assistance systems telling us when to stop driving as soon as they conclude that we are tired, aiming at preventing us from an accident. Mobile applications support our daily workouts and motivate us at sport activities, screens of our notebooks are automatically dimmed adjusting to the environmental light circumstances, both aiming at preventing health problems. And, based on the huge amount of web data we are able to derive relevant information and most probably foresee any epidemic or disaster situation, preventing its worst consequences. Behind all these solutions, there are different methods and algorithms that enable the processing of the different data and information and respectively support the intelligent environmental situation recognition. And, most of these solutions include at some point mining of temporal information. 

Mining of temporal information includes different approaches for processing, analysis, and forecasting of the data and information given. When it comes to the situation recognition, the temporal aspect of the information is very important. 


This workshop aims at merging research works relevant to mining temporal information from any kind of data with a special focus on (but not limited to) the following methods:

- Trend Mining
- Temporal Data Mining
- Topic Mining and Mining of Users Posts
- Data-based Situation Recognition
- Intelligent Situation Recognition
- Wireless Data Mining
- Signal Detection and Analysis
- Complex Event Processing and Messaging Analytics
- Semantic Complex Event Processing
- Data Stream Processing
- Data Mining on Streaming Data
- Semantic Technologies for Situation Recognition