Habitat loss and fragmentation occur at unprecedented rates, especially in tropical countries where human activities have deforested or degraded around 80% of tropical rainforests. Tropical forest fragmentation is considered the main cause of extinction of tropical forest avifauna, yet the mechanisms by which fragmentation affects bird populations are poorly understood. The present study investigates the pattern of bird species distribution in a fragmented landscape in tropical southeastern Mexico and the relation of bird community and species distribution patterns to landscape and fragment characteristics. Area and isolation of forest fragments were the main determinant of species richness and abundance in fragments, especially for forest-dependent species. The significant effect of isolation on species persistence in forest fragments suggests that limitation of dispersal is one potential mechanism by which fragmentation affects species distribution in the landscape. To understand how fragmentation can affect bird dispersal in a fragmented landscape, the effect of open areas and corridors on movement patterns of forest birds was investigated. The evidence presented in this study supports the idea that bird movement is restricted by open areas, especially for forest-restricted birds. Forested corridors had a positive effect on movement rates of forest birds, potentially acting to preserve movement and dispersal processes, and ultimately species persistence, in heavily fragmented landscapes.

Over the next decade, 5G networks will become more and more prevalent in everyday life. This will provide solutions to current limitations by allowing access to bands previously unavailable to civilian communication networks. However, this also provides new challenges primarily for the military operations. Radar bands have traditionally operated primarily in the sub-6 GHz region. In the past, these bands were off limits to civilian communications. However, that changed when they were opened up in the 2010's. With these bands now being forced to co-exist with commercial users, military operators need systems to identify the signals within a spectrum environment. In this thesis, we extend current research in the area of signal identification by using previous work in the area to construct a deep learning-based classifier that is able to classify a signal as either as a communication waveform (Single-Carrier (SC), Single-Carrier Frequency Division Multiple Access (SC-FDMA), Orthogonal Frequency Division Multiplexing (OFDM), Amplitude Modulation (AM), Frequency Modulation (FM)) or a radar waveform (Linear Frequency Modulation (LFM) or Phase-coded). However, the downside to this method is that the classifier is based on the assumption that all possible signals within the spectrum environment are within the training dataset. To account for this, we have proposed a novel classifier design for detection of unknown signals outside of the training dataset. This two-classifier system forms an open-set recognition (OSR) system that is used to provide more situational awareness for operators. / M.S. / Over the next decade, next-generation communications will become prevalent in everyday life providing solutions to limitation previously experienced by older networks. However, this also brings about new challenges. Bands in the electromagnetic spectrum that were reserved for military use are now being opened up to commercial users. This means that military and civilian networks now have a challenge of co-existence that must be addressed. One way to address this is being aware of what signals are operating in the bands such as either communication signals, radar signals, or both. In this thesis, we will developed a system that can do that task of identifying a signal as one of five communication waveforms or two radar waveforms by using machine learning techniques. We also develop a new technique for identifying unknown signals that might be operating within these bands to further help military and civilian operators monitor the spectrum.

Military systems are often part of critical operations where unplanned downtime should be avoided at all costs. Using modern machine learning algorithms it could be possible to predict when, where, and at what time a fault is likely to occur which enables time for ordering replacement parts and scheduling maintenance. This thesis is a proof of concept study for anomaly detection in monitoring data, i.e., sensor data from a ground based radar system as an initial experiment to showcase predictive maintenance. The data in this thesis was generated by a Giraffe 4A during normal operation, i.e., no anomalous data with known failures was provided. The problem setting is originally an unsupervised machine learning problem since the data is unlabeled. Speculative binary labels are introduced (start-up state and steady state) to approximate a classification accuracy. The system is functioning correctly in both phases but the monitoring data looks differently. By showing that the two phases can be distinguished, it is possible to assume that anomalous data during break down can be detected as well.  Three different machine learning classifiers, i.e., two unsupervised classifiers, K-means clustering and isolation forest and one supervised classifier, logistic regression are evaluated on their ability to detect the start-up phase each time the system is turned on. The classifiers are evaluated graphically and based on their accuracy score. All three classifiers recognize a start up phase for at least four out of seven subsystems. By only analyzing their accuracy score it appears that logistic regression outperforms the other models. The collected results manifests the possibility to distinguish between start-up and steady state both in a supervised and unsupervised setting. To select the most suitable classifier, further experiments on larger data sets are necessary. / Militära system är ofta en del av kritiska operationer där oplanerade driftstopp bör undvikas till varje pris. Med hjälp av moderna maskininlärningsalgoritmer kan det vara möjligt att förutsäga när och var ett fel kommer att inträffa. Detta möjliggör tid för beställning av reservdelar och schemaläggning av underhåll. Denna uppsats är en konceptstudie för detektion av anomalier i övervakningsdata från ett markbaserat radarsystem som ett initialt experiment för att studera prediktivt underhåll. Datat som används i detta arbete kommer från en Saab Giraffe 4A radar under normal operativ drift, dvs. ingen avvikande data med kända brister tillhandahölls. Problemställningen är ursprungligen ett oövervakat maskininlärningsproblem eftersom datat saknar etiketter. Spekulativa binära etiketter introduceras (uppstart och stabil fas) för att uppskatta klassificeringsnoggrannhet. Systemet fungerar korrekt i båda faserna men övervakningsdatat ser annorlunda ut. Genom att visa att de två faserna kan urskiljas, kan man anta att avvikande data också går att detektera när fel uppstår.  Tre olika klassificeringsmetoder dvs. två oövervakade maskininlärningmodeller, K-means klustring och isolation forest samt en övervakad modell, logistisk regression utvärderas utifrån deras förmåga att upptäcka uppstartfasen varje gång systemet slås på. Metoderna utvärderas grafiskt och baserat på deras träffsäkerhet. Alla tre metoderna känner igen en startfas för minst fyra av sju delsystem. Genom att endast analysera deras noggrannhetspoäng, överträffar logistisk regression de andra modellerna. De insamlade resultaten demonstrerar möjligheten att skilja mellan uppstartfas och stabil fas, både i en övervakad och oövervakad miljö. För att välja den bästa metoden är det nödvändigt med ytterligare experiment på större datamängder.

Embedded system such as microcontrollers has become more powerful and cheaper during the past couple of years. This has led to more and more development of on-edge applications, one of which is anomaly detection using machine learning. This thesis investigates the ability to implement, deploy and run the unsupervised anomaly detection algorithm called Isolation Forest, and its modified version Mondrian Isolation Forest on a microcontroller. Both algorithms were successfully implemented and deployed. The regular Isolation Forest algorithm resulted in being able to function as an anomaly detection algorithm by using both data sets and streaming data. However, the Mondrian Isolation Forest was too resource hungry to be able to function as a proper anomaly detection application.

Anomaly detection is a field of study that is closely associated with machine learning and it is the process of finding irregularities in datasets. Developing and maintaining multiple machine learning models for anomaly detection takes time and can be an expensive task. One proposed solution is to combine all datasets and create a single model. This creates a heterogeneous dataset with a wide variation in its distribution, making it difficult to find anomalies in the dataset. The objective of this thesis is then to identify a framework that is suitable for anomaly detection in heterogeneous datasets. A selection of five methods were implemented in this project - 2 supervised learning approaches and 3 unsupervised learning approaches. These models are trained on 3 synthetic datasets that have been designed to be heterogeneous with an imbalance between the classes as anomalies are rare events. The performance of the models are evaluated with the AUC and the F1-score, aswell as observing the Precision-Recall Curve. The results makes it evident that anomaly detection in heterogeneous datasets is a challenging task. The best performing approach was with a random forest model where the class imbalance problem had been solved by generating synthetic samples of the anomaly class by implementing a generative adversarial network. / Anomalidetektering är ett studieområde som är starkt förknippat med maskininlärning och det kan beskrivas som processen att hitta avvikelser i datamängder. Att utveckla och underhålla flera maskininlärningsmodeller tar tid och kan vara kostsamt. Ett förslag för att lösa dessa problem är att kombinera alla dataset och skapa endast en modell. Detta leder till att datamängden blir heterogen i dess fördelning och gör det mer utmanande att skapa en modell som kan detektera anomalier. Syftet i denna tes är att identifiera ett ramverk som är lämpligt för anomalidetektering i heterogena datamängder. Ett urval av fem metoder tillämpades i detta projekt - 2 metoder inom övervakad inlärning och 3 metoder inom oövervakad inlärning. Dessa modeller är tränade på syntetiska datamängder som är framtagna så att de är heterogena i dess fördelning och har en urbalans mellan klasserna då anomalier är sällsynta händelser. Modellernas prestanda evalueras genom att beräkna dess AUC och F1-värde, samt observera Precision-Recall kurvan. Resultaten gör det tydligt att anomalidetektering i heterogena datamängder är ett utmanande uppdrag. Den model som presterade bäst var en random forest model där urbalansen mellan klasserna var omhändertagen genom att generera syntetiska observation av anomaliklassen med hjälp av en generativ advarserial network.

Anomalies in data can be of great importance as they often indicate faulty behaviour. Locating these can thus assist in finding the source of the issue. Isolation Forest, an unsupervised machine learning model used to detect anomalies, is evaluated against two other commonly used models. The data set used were log files from a company named Trimma. The log files contained information about different events that executed. Different types of event could differ in execution time. The models were then used to find logs where some event took longer than usual to execute. The feature created for the models was a percentual difference from the median of each job type. The comparison made on various data set sizes, using one feature, showed that Isolation Forest did not perform the best with regard to execution time among the models. Isolation Forest classified similar data points compared to the other models. However, the smallest classified anomaly differed a bit from the other models. This discrepancy was only seen in the smaller anomalies, the larger deviations were consistently classified as anomalies by all models.

This thesis aims to uncover anomalies in the data describing the performance behavior of a "robot controller" as measured by software metrics. The purpose of analyzing data is mainly to identify the changes that have resulted in different performance behaviors which we refer to as performance anomalies. To address this issue, two separate pre-processing approaches have been developed: one that adds the principal component to the data after cleaning steps and another that does not regard the principal component. Next, Isolation Forest is employed, which uses an ensemble of isolation trees for data points to segregate anomalies and generate scores that can be used to discover anomalies. Further, in order to detect anomalies, the highest distances matching cluster centroids are employed in the clustering procedure. These two data preparation methods, along with two anomaly detection algorithms, identified software builds that are very likely to be anomalies. According to an industrial evaluation conducted based on engineers’ domain knowledge, around 70% of the detected software builds as anomalous builds were successfully identified, indicating system variable deviations or software bugs.

Market manipulation has increased in line with the number of active players in the financialmarkets. The most common methods for monitoring financial markets are rule-based systems,which are limited to previous knowledge of market manipulation. This work was carried out incollaboration with the company Scila, which provides surveillance solutions for the financialmarkets.In this thesis, we will try to implement a complementary method to Scila's pre-existing rule-based systems to objectively detect outliers in all available data and present the result onsuspect transactions and customer behavior to an operator. Thus, the method needs to detectoutliers and show the operator why a particular market participant is considered an outlier. Theoutlier detection method needs to implement interpretability. This led us to the formulation of ourresearch question as: How can an outlier detection method be implemented as a tool for amarket surveillance operator to identify potential market manipulation outside Scila's rule-basedsystems?Two models, an outlier detection model Isolation Forest, and a feature importance model (MI-Local-DIFFI and its subset Path Length Indicator) were chosen to fulfill the purpose of the study.The study used three datasets, two synthetic datasets, one scattered and one clustered, andone dataset from Scila.The results show that Isolation Forest has an excellent ability to find outliers in the various datadistributions we investigated. We used a feature importance model to make Isolation Forest’sscoring of outliers interpretable. Our intention was that the feature importance model wouldspecify how important different features were in the process of an observation being defined asan outlier. Our results have a relatively high degree of interpretability for the scattered datasetbut worse for the clustered dataset. The Path Length Indicator achieved better performancethan MI-Local-DIFFI for both datasets. We noticed that the chosen feature importance model islimited by the process of how Isolation Forest isolates an outlier.

The increasing volume of smart electricity meter readings presents a challenge forelectricity providing companies in accurately validating and correcting the associated data. This thesis attempts to find a possible solution through the application ofunsupervised machine learning for detection of anomalous readings. Through thisapplication there is a possibility of reducing the amount of manual labor that is required each month to find which meters are necessary to investigate. A solution tothis problem could prove beneficial for both the companies and their customers. Itcould increase abnormalities detected and resolve any issues before having a significant impact. Two possible algorithms to detect anomalies within these meters areinvestigated. These algorithms are the Isolation Forest and a Autoencoder, wherethe autoencoder showed results within the expectations. The results shows a greatreduction of the manual labor that is required up to 96%.

Real-time water quality monitoring using automated systems with sensors is becoming increasingly common, which enables and demands timely identification of unexpected values. Technical issues create anomalies, which at the rate of incoming data can prevent the manual detection of problematic data. This thesis deals with the problem of anomaly detection for water quality data using machine learning and statistic learning approaches. Anomalies in data can cause serious problems in posterior analysis and lead to poor decisions or incorrect conclusions. Five time series anomaly detection techniques: local outlier factor (machine learning), isolation forest (machine learning), robust random cut forest (machine learning), seasonal hybrid extreme studentized deviate (statistic learning approach), and exponential moving average (statistic learning approach) have been analyzed. Extensive experimental analysis of those techniques have been performed on data sets collected from sensors deployed in a wastewater treatment plant. The results are very promising. In the experiments, three approaches successfully detected anomalies in the ammonia data set. With the temperature data set, the local outlier factor successfully detected all twenty-six outliers whereas the seasonal hybrid extreme studentized deviate only detected one anomaly point. The exponential moving average identified ten time ranges with anomalies. Eight of them cover a total of fourteen anomalies. The reproducible experiments demonstrate that local outlier factor is a feasible approach for detecting anomalies in water quality data. Isolation forest and robust random cut forest also rate high anomaly scores for the anomalies. The result of the primary experiment confirms that local outlier factor is much faster than isolation forest, robust random cut forest, seasonal hybrid extreme studentized deviate and exponential moving average. / Thesis / Master of Computer Science (MCS)