data mining

Slide 1: 

Presentation On Neural Networks Submitted To: Presented By: Lec.Shraddha Masih Nandkishor Buwade Avinash Giri Vijay Kumar Chourase

Slide 2: 

What is Neural Network. Neural network are a different paradigm for computing , which draws its inspiration from neuroscience . The brain consist of a network of neurons , each of which is made up of a nerve fibres called dendrites , connected to the cell body where the cell nucleus is located.

Slide 3: 

Evolution Of Neural Networks. The evolution of neural network as a new computation model originates from the pioneering work of McCulloch and Pitts in 1943. They suggest a simple model of a neuron that commuted the weight sum of the inputs to the neuron and an output of 1 or 0. A zero output would corresponds to the inhibitory state of the neuron . One would corresponds to the excitatory state of the neuron.

Slide 4: 

A Simple Perceptron : The network has 2 binary inputs , I0 and I1 and output Y. W0 and W1 are the connection strengths of input 1 and input 2. i0 i1 Bias unit Linear threshold unit Y=i0 or i1 W0 W1 Wb Wb is threshold

Slide 5: 

Multi- Layer Perception (MLP) : MLP is a development from the simple perception in which extra hidden layers are added. More then one hidden layer can be used .The network topology is constrained to be feed forward. Generally ,connection are allowed from the input layer to the first hidden layer , from the first hidden layer to the second layer and so on until the last hidden layer to the output layer . Input layer Hidden layer Out put layer

Slide 6: 

Radical Basis Function Network : Radical basis function (RBF) network are a also feed forward but have only one hidden layer. RBF hidden layer units have a receptive field which has a centre. Input layer Hidden layer Output layer RBF Architecture

Slide 7: 

Perceptron Learning Rule : This is the first learning scheme of neural computing . The weight are changed by an amount proportional to the Difference between the desired output and actual output. The formula is given by Wi=L(D-Y). Ii Where L is learning rate. D is the desired output. Y is the actual output. Single perceptrons are limited the simple hyperplane decision surface.

Slide 8: 

Training In MLP : For the nodes in the output layer , it is easy to compute the error as we know the actual outcome and the desired result . For the nodes in the hidden layers , since we do not know the desired result. We propagate the error computed in the last layer backward .this process gives the change in the weight for the edges layer wise. This stander method used in training MLP s is called the back propagation algorithm . The learning steps consist of the Forward pass : The out put and the error at the output units are calculated. Back ward pass : The output unit error is used to alter weights on the out put units

Slide 9: 

KOHONEN’ S SOM : The self-organizing map (SOM) was a neural network model developed by TEUVO KOHONEN during 1979-82. SOM is one of most widely used unsurprised NN model and employs competitive learning steps. It consists of a layer of input units , each of which is fully connected to a set of output units. These output units are arranged in some topology most common choice is a two dimensional grid. SOM architecture 2 dimensional array of output units High dimension input X

Slide 10: 

Application of neural networks : investment analysis : to predict the moment of stock ,currencies etc. from previous data . There , they are replacing earlier models . Monitoring : network have been used to monitor of aircraft engine . By monitoring vibration levels and sound , an early warning of engine problem can be given. marketing : neural network have been used to improve marketing mailshots.

Slide 11: 

Data Mining using NN : A Case Study 10 9 8 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8 9 10 Population Clusters

Slide 12: 

Knowledge Extraction Though Data Mining : Kohonen, self –organizing maps (SOM) are used to cluster a specific medical data set containing information about the patient’s drug, topographies and morphologies. Morphologic Tree Condensed : Level 1 level2 0 General morphologic no level 2 term codes terms 1 Traumatic Abnormalities 10 General & compression injuries 11 Thermal cold & irradiation injuries 12 Fractures 13 Dislocation & ankyloses 14 Wounds 15 Implants & transplants 16 Acquired Absences 17 Amputated & transected structures 18 Operative sites

Slide 13: 

2 Congential Abnormalities no level 2 term codes 3 Mechanical Abnormalities 4 Types of inflameation 30 Calculi & Foreign Bodies 31 Displacement & Deformities 32 Dilations & Diverticula's 33 Retention & cysts 34 Obstruction & stenoses 35 Blood clots ,thrombi & emboli 36 Fluid disturbances 37 Hemorrhages 38 ulcers 41 General Terms 42 Acute inflammation 43 Subacute inflammations 44 Granulamation inflammation 45 Organizing inflammation

Genetic Algorithms: : 

Genetic Algorithms: A Genetic Algorithm (GA) is a computational model consisting of five parts: A starting set of individuals, P. Crossover: technique to combine two parents to create offspring. Mutation: randomly change an individual. Fitness: determine the best individuals. Algorithm which applies the crossover and mutation techniques to P iteratively using the fitness function to determine the best individuals in P to keep.

Life cycle of Algorithm: : 

Life cycle of Algorithm: Creation of a population of strings. Evaluation of each string. Selection of the best strings. Genetic manipulation to create a new population of strings.

Genetic Algorithm : 

Genetic Algorithm

GA Advantages/Disadvantages : 

GA Advantages/Disadvantages Advantages Easily parallelized Disadvantages Difficult to understand and explain to end users. Abstraction of the problem and method to represent individuals is quite difficult. Determining fitness function is difficult. Determining how to perform crossover and mutation is difficult.

Data Mining using GA : 

Data Mining using GA The application of GA in the context of data mining is generally for the task of hypothesis testing & refinement. Hypothesis refinement is achieved by “seeding”. The important aspect of the GA app. is the encoding of the hypothesis & the evaluation function for fitness. The genetic algorithm can be used for optimal decision tree induction.

Rough Sets: : 

Rough Sets: Rough set theory is a tool for studying imprecision, vagueness & uncertainty in data analysis. It focuses on delivery patterns, rules & knowledge in data. The rough set is the approximation of a vague set by pair of precise concept called the lower & upper approximation.

Applications of Rough Set: : 

Applications of Rough Set: RSES: The system was developed in Poland. A maximum of 30,000 objects with 16,000 attributes can be processed for rule generation. KDD-R: This system is based on the Variable Precision Rough Set model. -- Analysis of dependencies among attributes & elimination of superfluous attributes. -- Computation of rules from data. LERS (Learning from examples based on rough sets) -- The inconsistencies are not corrected but the upper & lower approximation for each concept is calculated.

Support Vector Machines : 

Support Vector Machines Support vector machines are learning machines that can perform binary classification and regression estimation tasks. SVMs minimize the expected error rather than minimizing the classification error SVMs employ the expected error duality theory of mathematical programming to get a dual problem that admit efficient computational methods.

Conti….. : 

Conti….. There are two results make this approach successful The generalization ability of this learning machine depends on the VC dimension of the set of functions that the machine implements rather than on the dimensionality of the space. The construction of the classifier only needs to evaluate an inner product between two vector of the training data.

Conclusion : 

Conclusion We have discussed four different techniques Neural Networks, Genetic Algorithm, Rough Set Theory and Support Vector Machines. These techniques are helpful for developing decision tree, clustering and discovering the association rules in the data mining. These techniques are undoubtedly important and efficient techniques for data mining applications.

Thank You ! : 

Thank You !