Comparative Analysis of Data Mining Techniques on Educational Data set

Data mining can be described as a process of analyzing hidden patterns of data according to different perspectives for categorization into useful information, which is collected and assembled in common areas, such as data warehouses, for efficient analysis, data mining algorithms, facilitating business decision making and other information requirements to ultimately cut costs and increase revenue.

Following are the major steps of data mining process.

• Extract, transform and load data into a data warehouse.
• Store and manage data in a multidimensional databases.
• Provide data access to business analysts using application software.
• Present analyzed data in easily understandable forms, such as graphs.

Referring to a journal written by Sumit Garg and Arvind K.Sharma, data mining is a relatively young and interdisciplinary field of computer science which discovers new patterns in large data sets. Data mining is a wide research area with important applications in Engineering, Science, Medicine, Business and Education.

Data Mining Techniques

According to the journal, as data mining process is to find new pattern from large data sets, it has several techniques. They can be categorized into two, as supervised learning and unsupervised learning.

Classification : 
Used to develop a model which can classify the population of records at large level.

Decision Tree : 

It’s a structure like a flowchart where each node denotes test on an attribute value, each branch represents the results of the test, while tree leaves show classes.

Bayesian Classification : 

It’s a statistical classifier. Can be used to predict probabilities of class membership. This is based on Bayes Theorem which is,

Neural Networks :

Belongs to the border line between the artificial intelligence and approximation algorithm. It’s a collection of neurons like processing units with weighted connection between the units.

Association Rule Mining : 

It’s an important data mining model studied extensively by the database and data mining community. It is, finding frequent patterns, associations, correlations, objects in transactional databases, relational databases.

Clustering : 

Process of partitioning a set of data in a set of meaningful sub - classes, called clusters. It helps to understand the natural grouping or structure in a data set.

Prediction :

Used to identify the relationship between independent variables and relationship between dependent and independent variables.

Time Series Analysis :

It’s a sequence of data points, measured typically at successive times spaced at uniform time intervals.

Sequential Patterns :

Seeks to discover similar patterns in data transaction over a business period.

The authors of this journal have proposed a methodology which used to generate a database for their study.

A data set which represents their data-set collected from an Engineering College, as following :  

Their experiments and observations was done by WEKA data mining tool. WEKA (Waikato Environment for Knowledge Analysis) is developed at the University of Waikato, New Zealand. It’s implemented in Java programming language and has GUI for loading data, running analysis and producing visualization of result. WEKA supports classification, clustering, feature section, data preprocessing and regression, visualization as well.

WEKA GUI chooser provides interfaces to work on Explorer, Experimenter, Knowledge Flow and Simple CLI.

Comparisons

In their journal, they have well described the importance of data mining techniques and comparisons of data mining techniques.

Reference : 

Sumit Garg, M.Tech Scholar, Dept. of Computer Science, Shekhawati Engineering College, Dundlod, Rajasthan, India, Arvind K. Sharma, Guest Faculty, Dept. of Computer Science, University of Kota, Kota, Rajasthan, India, International Journal of Computer Applications (0975 – 8887), Volume 74– No.5, July 2013

Vehicle detection and traffic information extracting in CCTV streams.

https://www.researchgate.net/publication/44113796_Vehicle_localisation_and_classification_in_urban_CCTV_streams

This paper presents a suitable framework for vehicle detection and localization that introduced as part of a project at Transport for London (TfL). In this research project, two main vehicle detection methods are highlighted. 

• Motion silhouettes
• 3DHOG (3D extended Histograms of oriented gradients)

The researches understood the fact that the traffic flow monitoring by detecting vehicles in CCTV streams tend to be unique and accurate by using high camera numbers to track the long transmission paths. In fact the research shown that manual monitoring over time significantly reduces the accuracy of detection. Therefore they used new technology that provides automatic and relevant real – time alerts.

The conventional concept of these types of projects is background modelling. In the motion mask method, it is only used as baseline. Motion mask cannot fulfill the vehicle identification process properly due to some environmental reasons like lightning changes, rain and fog. Camera shake and low camera angles can also effect to that process partially. After realizing that problem a traffic based classifier 3DHOG was introduced to overcome the problems.

The framework defined for vehicle detection in this research contains three basic components.

          i.  Detector

         ii.  Classifier

         iii. Tracker

Detector is designed to model the static part of the scene. It uses background estimation with Gaussian Mixture Model (GMM). This model enables rejection of moving objects in the background.

The difference between the background and the new frame, helps to estimate the foreground mask. It is refined by a shadow removal algorithm. The contours of the resulting binary foreground mask are extracted. The maximum match of vehicle gives the final estimation.

The motion silhouette method cannot overcome the problems of shadow removing, noise and saturation areas of the cameras. 3DHOG (Histograms of Oriented Gradients) uses to substitute overlapping match measures of the frames. HOG descriptor to image patches defined in 3D model space.

Affine transformation used to generate those scale normalized patches. Every patch generated a descriptor. In each descriptor is generated from every patch. Using the data driven appearance model, the single Gaussian distribution is learned for every interest point descriptor.  During system operation, new descriptors are generated for every 2D projection blocks. The remainder of the framework remains identical to the earlier description.

As the last step, the frame to frame detection is used as inputs to the Kalman filters. The detection of a single frame can be tracked over time using a Kalman filter.

Vehicle detection result, using 3DHOG:

A Case Study for Traffic Control Signal at Four Way Intersection Road

The research article “A Case Study for Traffic Control Signal at Four-Way Intersection Road” was written by four people, Kamlesh Kumar Pandey, Rajat Kumar Yadu, Pradeep Kumar Shukla and Narendra Pradhan. The authors of the article have focused on coming up with a solution which has the capability of overcoming one of the major problems in the world, the traffic management. They have pointed out the important fact regarding the difficulty that a single person faces when handling traffic congestion in a multi-lane intersection. Throughout the article, their goal is to describe a mechanism to handle large traffic flows in four way junctions using fuzzy logic. According to the authors, fuzzy logic technology possesses the capability of mimicking the human intelligence for controlling traffic lights.

They expect to use two different types of traffic controlling methods for the intersection where the first method is to use a preset cycle time to change the light patterns. The other type is to combine the preset cycle time with a sensor in order to dynamically change the light patterns based on fuzzy logic. The structural diagram of their system is given below.

Figure 01 - Structure of the System

They use two sensors to gather information of the vehicles where one of them is responsible for counting the number of vehicles passing the traffic lights while the other sensor counts the number of vehicles coming towards the intersection. The fuzzy logic controller too functions as an important part of the system as it controls the length of the green light time according to the traffic condition.

Figure 02 - Structure of the Fuzzy Controller

The fuzzy logic controller determines whether to extend or terminated the currently implemented light pattern (i.e green light time). The degree of priority that should be allocated to each lane is analyzed by the set of fuzzy rules and relevant decisions are made based accordingly. The fuzzy controller design is said to contain several functions that include Fuzzification, Fuzzy Decision Making, Defuzzification and Extension Time.

Fuzzification

According to the authors, the process of fuzzification involves converting crisp set data into a fuzzy set. The number of vehicles counted by the sensors, is sent to the system that facilitates fuzzification. The proposed fuzzy set supports a numerical number range between 0 and 1.  

Fuzzy Decision Making 

The fuzzy decision making is a process that takes the values from the previously converted data to make decisions using the knowledge based and the fuzzy rule base which are integrated within the decision making system. The knowledge base functions as subsystem to determine the best possible pattern that can be implemented using the input variables. The knowledge base is said to be constructed using the results obtained by traffic simulation. Further, the functions of knowledge base are complemented by rules and searching algorithms such as heuristics and hill clamming search.

Defuzzification

Defuzzification is the process that converts fuzzy output values back to the crisp values. The output value of this process is related to how long the green light should be lit.

The authors have tried to come up with a really good solution through the use of fuzzy logic to overcome the traffic congestion problem. However, the research article contains a lot of ambiguous content since the language that is used throughout the article cannot be considered good enough for an article. 

Reference:

Kamlesh Kumar Pandey, Rajat Kumar Yadu, Pradeep Kumar Shukla3, Narendra Pradhan, "A Case Study for Traffic Control Signal at FourWay Intersection Road" International Journal of Computer Techniques -– Volume 2 Issue 4, July - Aug 2015, ISSN: 2394-2231

  

Image Processing Based Traffic Light Control

                Traffic lights are most important part in modern roads. But there some problems in the traffic lights. Green light for empty roads and red light for maximum vehicles present on the road are the major problems in traffic lights. So many people try to solve these problems. As a result of it Omkar Ramdas Gaikwad, Anil  Vishwasrao,   Prof. Kanchan Pujari and  Tejas Talathi found a solution for it. They introduced a Image Processing Based Traffic Light Controlar to the world.

              In this project they have used video camera. Camera is shooting video and the video is then converted to sequence of images by taking snapshots. By taking images we can get a basic idea about the real traffic condition. In this technique they would be drawing lines of green colour at an arbitrary distance in the road.  By using that lines they can calculate the traffic density of the road. In here the camera would be placed at the signal which would be focusing on the line drawn. In this project they used the MATLAB software and the entire algorithm for image processing is implemented in MATLAB.

                First they get the real time traffic image and converted it in to grayscale. After that green colour subtraction is done for measuring pixel count. Then they converted this image to black and white format. As a result of it the green strip area is in white colour and background is in black. Then they measured the white pixel area of the image. If  the white area pixel is very low the traffic is very high. As well as the white pixel area is very high the traffic is very low. After getting the pixel count they generate the final result.

               If the traffic is very high the green light active for 35 seconds. If the traffic is medium the green light active for 30 seconds. If the traffic is very low the green light active for 25 seconds. If there is no traffic red light active.

                I think this is a good technique to control the traffic lights. We can use these decision making models and algorithms for our project also. Then we can build a better technique to control the traffic lights.

• References :-

Omkar Ramdas Gaikwad, Anil Vishwasrao, Prof. Kanchan Pujari, Tejas Talathi, "Image Processing Based Traffic Light Control" , International Journal of Science, Engineering and Technology Research (IJSETR) , Volume 3, Issue 4, April 2014

Research on Data Mining Models for the Internet of Things

Data mining is the process of sorting large data sets to identify patterns and establish relationships for solving problems through data analysis. A data mining model is created by applying an algorithm to data.

Some students of Zhejiang University have done a research on Data Mining Models for the Internet of Things (IoT). They have defined the IoT as the next generation of internet which contains trillions of nodes representing various objects from small ubiquitous sensor devices and handhelds to large web servers and super computer clusters.

They have mentioned four data models for the Internet of Things.

• Multi-layer data mining models.
• Distributed data mining model.
• Grid based data mining model.
• Data mining model from multi-technology integration perspective.

Multi-layer data mining model – They have shown four layers for the multi-layer data mining model.

• Data collection layer
• Data management layer
• Event processing layer
• Data mining service layer

Data collection layer adopts devices to collect various smart object’s data.

Data management layer applies centralized or distributed database or data warehouse to manage collected data.

Event processing layer analyses events in IoT effectively.

Data mining service layer is built based on data management and event processing. It is self-oriented.

Distributed data mining model for IoT solves the problems brought by distributed storage of nodes and decomposes the complex problems into simple problems.

Grid based data mining model supports to implement the data mining operations for IoT.

Data mining model for IoT from multi-technology integration perspective is shown in the following figure.

They have given six key issues in data mining of IoT.

They are,

• Data collection from smart objects of IoT - When conducting data collection from smart object of IoT, the special needs of smart objects should be taken into account.
• Data abstraction, compression, index, aggregation and multi-dimensional query – Since the internet of things produces a massive data of smart objects. It is better to consider how to manage data of IoT effectively and how to implement online analytical query and processing conveniently.
• Event filtering, aggregation and detection – Event filtering and complex event processing process simple events in data. First, data are aggregated accordingly to events. While primitive events are filtered, valuable events are obtained. Then the simple events are integrated into complex events.
• Centralized data processing and mining Vs. Distributed data processing and mining – centralized or distributed data processing and mining models can be adopted flexibility in different situations.
• Research on data mining algorithms for IoT – A key issue is to study the novel data mining algorithms for IoT. The main tasks are classification, forecasting, clustering, outlier detection, association analysis, spatial and temporal patterns mining for IoT.
• Data mining towards the next generation of internet – The next generation of internet has many potential direction of development such as IPv6, ubiquitous networks, trusted network, semantic web, grid, service oriented applications, optical transmission, cloud computing etc. These technologies will integrate with IoT. In that case, many new data mining problems need to be studied intensively.

In a nut shell, in their research paper, they have defined what data mining models are and what IoT means. Also they have given four models of data mining and among them, the multi-layer model has four layers. Then they have discussed the key issues in data mining of IoT.

Reference:

Shen Bin, Liu Yuan, Wang Xiaoyi, Ningbo Institute of Technology, Zhejiang University, Ningbo, China, College of Management, Zhejiang University, Hangzhou, China. “Research on Data Mining Models for the Internet of Things”

A Video-based Real-time Vehicle Detection Method by Classified Background Learning

Research by Xiao-jun Tan, Jun Li & Chunlu Liu: Link to Research Article


In this research, they propose a new two-level real-time vehicle detection method in order to meet the robustness and efficiency requirements of real world applications. At the high level, pixels of the background image are classified into three categories according to the characteristics of Red, Green, and Blue (RGB) curves. The robustness of the classification is further enhanced by using line detection and pattern connectivity. At the lower level, an exponential forgetting algorithm with adaptive parameters for different categories is utilised to calculate the background and reduce the distortion by the small motion of video cameras.


Traffic Scene Analysis and Classification

Pixels in the traffic scene are classified into three categories according to the RGB curves, namely road surface, lane line and others. They found two common properties from RGB curves of a road surface pixel (figure a) and lane line (figure b).

 

1. The differences of three primary colours (Red, Green and Blue) are not significant.
2. The range of the oscillations is small.  

However, pixels in different categories have different average values. It is found that the road surface pixel has an average value smaller than 150 while the lane line pixel has an average value greater than 180.

By using equations, they have shown that a pixel can be classified into 3 categories.

1. A pixel is on the road surface
2. A pixel is on the lane line
3. The pixels are neither on the road surface nor on the lane line

After the classification, a procedure is added to enhance the robustness of the classification, which is based on the observations of the geometric characteristics of the highway.

Line Detection: With the results of the classifier, the candidates of a lane line form a straight line, which can be detected using a Hough Transformation. Only those near the line can be confirmed as lane lines. The detected line is in colour blue. Two white areas due to plastic bags are excluded. If the lane line is another curve than a straight line, then other techniques can be adopted.

Pattern Connectivity: A growing algorithm, similar to that from Tremeau and Borel, is utilised to check the candidates of the road surface. The result is shown in green in figure below. In figure below, the green pixels are the connected patterns of road surfaces. Those pixels that cannot be classified either as road surface or as a lane line are labelled with their original colour.

Background Learning based on Classification 

The background learning algorithm is similar to the exponential forgetting algorithm that has been mentioned in the research article previously. But in here they have changed the forgetting factor to a function rather than a constant. When compared with the simple forgetting algorithm, the modified forgetting factor considers the motion happening at the current pixel so that only the relatively stable pixels are chosen to perform background learning. This approach can significantly reduce the effects caused by small motion of video cameras, which is very important for outdoor environments. If the pixel belongs to the other class, then background learning will not be performed because this kind of pixel has nothing to do with the motion detection of the current frame. To further reduce the computation time, the background classification and background learning are not updated synchronously. As shown in the figure below, background learning is performed for almost every frame, but the classification will only be carried out after a certain period.

 

Vehicle Detection 

Vehicle detection has almost the same meaning of motion segmentation, which judges whether there is a motion at a certain pixel. In the proposed method, the area to detect vehicles is not the whole image, but rather due to understanding given at the higher level. If the pixel belongs to the other class and is in the lane line or near the lane line, then detection will not be performed. This can help to avoid false detection due to small motions of the camera. Figure below shows each area with a possible vehicle, while other areas are ignored. An algorithm named Minimum Boundary Rectangles (MBRs) is then utilised in order to divide those areas that belong to different vehicles. The rectangles indicating the moving vehicles are outlined.