BitVAna: A Solution for Clustering and Analysing Unstructured Data in Real-time
 
The goal of this research project is to develop an end-to-end solution which can perform collecting, curating, analyzing in a sequence to predict events related to logistics. So far several tasks have been done to achieve the goal. A volume of literature about technologies and methods such as machine learning methods were studied during literature review. Also, we carried out a feasibility study with existing tools. We developed a solution that fetches data from Twitter, Facebook, weather sensors, and online newsfeeds, curate and wrangled them in CSV (comma separated variable) format, and classify them using our model.
Contact us Contact us to more about our solution!

clustering

Fig. – The High-level Architecture of BitVAna.


MINERVA: A Deep Learning Based Model for Predictive Trading Analytics
 
We developed a model for predicting stock prices using deep learning techniques. We implemented the model by using Big Data technologies. Our solution has two integral parts: learning and extracting features and predicting the stock price based on the learned features. We studied different types of deep learning techniques which include Deep Belief Network (DBN), Recurrent Neural Network (RNN) for building predictive model. In addition, there are Convolution Neural Network (CNN), Autoencoder, and Boltzman Machine for feature learning. Our study reveals that CNN is the most potential candidate for our solution. We developed a CNN based predictive model which enables performing analysis on batch dataset. We implemented the model using Python on Google Cloud Platform. It is worth noting that we reused an existing model proposed by Google. We experimented our model with dataset collected from nine different stock exchanges such as NYSE, NASDAQ, HSE etc.

clustering

Fig. – The High-level Architecture of MINERVA.


InSFram: An Integrated Scalable Framework and a Model for Logistics Business Process Analytics to Optimize Routing in Realtime
 
In this research project, we developed a scalable framework which can handle massive-scale data and enables integrating data from different sources. Additionally, we aimed to design and develop model which uses streaming data and produces an alternative route in realtime to prevent excessive delay. We developed a framework which we named InsFrame by marrying technologies from two different domains: business process and Big Data. We bundled process design and management tool Activiti with Apache Hadoop, Apache Kafka, and MongDB, and Apache HBase. The framework integrates shipment management process logs with data streams sourcing from Lyon Smart City. In addition, we developed a model to visualize traffic, availability of biking space. We plan to develop a model that produces an alternative routing in realtime.

clustering

Fig. – The High-level Architecture of InsFram.


SANA: A Context-Aware Service-based system for sentiment analysis and visualisation of Twitter data in Realtime
 
SANA is a service-based solution for analysing data to extract sentiments of “things” (e.g. products, service, and other real-world entities) in realtime.The analysis heavily relies on the contexts of data. The contexts are essentially the topics of tweets such as location, person, organization etc. SANA consists of a list of components which are: data collector, data processor, storage, and visualizer. These components perform different tasks including fetching and ingesting data, analyse them, storing, and visualising results.

clustering

Fig. – The Demonstration of SANA.