Information System and Artificial Intelligence

1

An Information System (IS) collects, processes, stores, and distributes information to support decision-making. It combines hardware, software, data, people, and procedures.

2

The CIA Triad is the foundation of IS security: Confidentiality (only authorized access), Integrity (data is accurate and unaltered), Availability (data is accessible when needed).

3

IS Controls are safeguards: Preventive controls stop incidents before they happen (firewalls, access control), Detective controls identify incidents during or after (audit logs, intrusion detection), Corrective controls restore after an incident (backup recovery, patches).

4

Authentication verifies WHO you are (passwords, biometrics). Authorization determines WHAT you can do (permissions, roles). Authentication always comes before authorization.

5

Multi-Factor Authentication (MFA) uses two or more of: something you know (password), something you have (OTP token), something you are (fingerprint). MFA drastically reduces account compromise risk.

6

SSL/TLS encrypts data in transit between client and server. SSL (Secure Sockets Layer) is deprecated; TLS (Transport Layer Security) is the current standard. Both use asymmetric key exchange then symmetric encryption.

7

A Certificate Authority (CA) is a trusted third party that issues digital certificates binding a public key to an identity. Browsers trust certificates signed by recognized CAs.

8

IS Audit evaluates controls, security, and compliance of an information system. It checks whether IS controls adequately protect data and align with organizational policies.

9

Risk Management in IS: Risk = Threat × Vulnerability × Asset Value. Strategies: accept (low risk), mitigate (add controls), transfer (insurance), avoid (stop the activity).

10

Data integrity mechanisms: Checksums detect errors, Hash functions (MD5, SHA-256) verify data hasn't changed, Digital signatures verify both integrity and origin.

1

Enterprise Resource Planning (ERP) integrates internal business processes — finance, HR, manufacturing, supply chain — into a single system. Key vendors: SAP, Oracle. All departments share one database.

2

Supply Chain Management (SCM) manages the flow of goods and services from supplier → manufacturer → distributor → retailer → customer. Goal: reduce costs, improve efficiency, optimize inventory.

3

Customer Relationship Management (CRM) manages customer-facing interactions — sales, marketing, customer service. Goal: improve customer satisfaction and retention. Examples: Salesforce, HubSpot.

4

OLTP (Online Transaction Processing) handles day-to-day operational transactions: inserting, updating, deleting records. Optimized for high-frequency, short transactions with ACID properties. Example: banking ATM, POS.

5

OLAP (Online Analytical Processing) handles complex queries for decision support. Optimized for read-heavy, large-scale aggregations. OLAP uses multidimensional data cubes. Example: sales trend analysis.

6

A Data Warehouse is a subject-oriented, integrated, time-variant, non-volatile collection of data supporting management decision-making. The 4 characteristics (Inmon's definition): Subject-oriented (organized by subject like sales), Integrated (consistent formats), Time-variant (stores historical snapshots), Non-volatile (data is only loaded, not updated).

7

Data Mining extracts hidden patterns and knowledge from large databases. Techniques: classification, clustering, association rules, regression. Applications: market basket analysis, fraud detection.

8

Knowledge Management (KM): Tacit knowledge is personal, experience-based, hard to document (skills, intuitions). Explicit knowledge is codified, documented, easily shared (manuals, databases). KM systems aim to convert tacit to explicit.

9

Decision Support System (DSS) helps managers make semi-structured or unstructured decisions. Components: database, model base, user interface. Different from EIS (Executive IS) which serves top management.

10

Business Intelligence (BI) transforms raw data into actionable insights using OLAP, data warehousing, reporting dashboards, and data mining. Helps organizations make data-driven decisions.

1

The Balanced Scorecard (BSC) by Kaplan & Norton aligns IS strategy with business goals through 4 perspectives: Financial (profitability), Customer (satisfaction, retention), Internal Business Process (efficiency), Learning & Growth (employee skills, innovation).

2

Cloud Computing service models: IaaS (Infrastructure as a Service — rent VMs, storage; e.g., AWS EC2), PaaS (Platform as a Service — deploy apps without managing OS; e.g., Heroku, Google App Engine), SaaS (Software as a Service — use software over internet; e.g., Gmail, Salesforce). IaaS > PaaS > SaaS in control level.

3

Cloud deployment models: Public cloud (shared, owned by provider), Private cloud (dedicated, on-premises or hosted), Hybrid cloud (mix of both), Community cloud (shared by organizations with common concerns).

4

MapReduce is a programming model for processing large datasets in parallel across a cluster. Phase 1: Map — splits input, applies function to each chunk producing key-value pairs. Phase 2: Shuffle/Sort — groups by key. Phase 3: Reduce — aggregates values per key. Used in Hadoop.

5

Hadoop is an open-source framework for distributed storage and processing of big data. Core components: HDFS (Hadoop Distributed File System — stores data across nodes with replication), YARN (resource manager), MapReduce (processing engine). Default HDFS replication factor: 3.

6

Collaborative Filtering (used in recommendation systems) recommends items based on what similar users liked (user-user) or what similar items were liked by (item-item). Example: Netflix 'users like you also watched...'

7

Content-Based Filtering recommends items similar to what a user has previously liked, based on item features. Example: if you liked Action movies, recommend more Action movies.

8

Systems Development Life Cycle (SDLC): Planning → Analysis → Design → Implementation → Testing → Deployment → Maintenance. Waterfall is sequential; Agile is iterative.

9

E-commerce models: B2B (business to business), B2C (business to consumer), C2C (consumer to consumer e.g., eBay), G2C (government to citizen). E-commerce requires secure payment gateways, SSL/TLS, and digital certificates.

10

Big Data is characterized by the 5 Vs: Volume (massive scale), Velocity (speed of generation), Variety (structured/unstructured/semi-structured), Veracity (data quality), Value (actionable insights).

1

Artificial Intelligence (AI) is the simulation of human intelligence by machines. Goals: learning, reasoning, problem-solving, perception, language understanding.

2

An Intelligent Agent perceives its environment through sensors and acts on it through actuators. Types: Simple reflex, Model-based reflex, Goal-based, Utility-based, Learning agent.

3

Search algorithms: Uninformed (Blind) search has no domain knowledge — BFS (complete, optimal for unit cost), DFS (not optimal, not complete), Iterative Deepening DFS (IDDFS), Uniform Cost Search. Informed (Heuristic) search uses domain knowledge — Greedy Best-First, A*.

4

**A* Search** is the optimal informed search algorithm. Evaluation function: f(n) = g(n) + h(n) where g(n) = cost from start to node n, h(n) = estimated cost from n to goal (heuristic). A* is optimal if h(n) is admissible (never overestimates).

5

Minimax algorithm is used for two-player zero-sum games (e.g., Chess, Tic-Tac-Toe). MAX player maximizes score; MIN player minimizes it. Minimax searches the entire game tree.

6

Alpha-Beta Pruning is an optimization of Minimax that prunes branches that cannot influence the final decision. Alpha (α) = best MAX can guarantee; Beta (β) = best MIN can guarantee. Prune when α ≥ β. Can reduce time complexity from O(b^d) to O(b^(d/2)).

7

Machine Learning (ML) types: Supervised (labeled training data, learns input→output mapping — classification/regression), Unsupervised (no labels, finds structure — clustering/dimensionality reduction), Reinforcement Learning (agent learns by reward/penalty from environment).

8

Overfitting = model learns training data too well (memorizes noise), poor generalization. Underfitting = model too simple, misses patterns in training data. Solution: cross-validation, regularization, more data.

9

k-Nearest Neighbors (kNN) classifies by finding k nearest training examples to new point and taking majority vote. Lazy learner (no explicit training phase). Distance metric: usually Euclidean. Sensitive to irrelevant features and scale.

10

Natural Language Processing (NLP) enables computers to understand human language. Key tasks: tokenization, parsing, sentiment analysis, machine translation, named entity recognition. Libraries: NLTK, spaCy.

1

A Perceptron is the simplest neural network — a single layer with inputs, weights, bias, and a step activation function. It can only solve linearly separable problems. Cannot solve XOR without hidden layers.

2

Multi-Layer Perceptron (MLP) adds one or more hidden layers, enabling non-linear decision boundaries. This overcomes the perceptron's limitation and can solve XOR, image classification, etc.

3

Backpropagation trains neural networks by computing the gradient of the loss with respect to each weight using the chain rule, then updating weights with gradient descent. Flow: Forward pass (compute output) → Compute loss → Backward pass (compute gradients) → Update weights.

4

Activation functions: Sigmoid (0 to 1, used in binary classification output), ReLU (max(0,x), avoids vanishing gradient, most popular in hidden layers), Tanh (-1 to 1), Softmax (multi-class output, probabilities sum to 1).

5

Bayesian Network is a Directed Acyclic Graph (DAG) where nodes represent random variables and directed edges represent conditional dependencies. Used for probabilistic inference. P(A,B,C) = P(A) × P(B|A) × P(C|A,B).

6

Markov Network (Markov Random Field) uses an undirected graph to represent symmetric, non-causal dependencies between variables. Unlike Bayesian networks, no direction is implied. Used in image segmentation.

7

Hidden Markov Model (HMM) models sequences where the state is hidden (unobserved) but produces observable output. Key components: states, observations, transition probabilities, emission probabilities. Applications: speech recognition, DNA analysis.

8

Genetic Algorithm (GA) mimics biological evolution: Selection (choose best individuals based on fitness), Crossover (combine two parents to create offspring), Mutation (randomly alter genes). The fitness function evaluates solution quality.

9

Deep Learning uses neural networks with many layers (deep architectures). CNN (Convolutional Neural Networks) excels at image processing; RNN (Recurrent Neural Networks) handles sequential data; LSTM handles long-term dependencies.

10

Vanishing Gradient Problem: in deep networks, gradients become very small in early layers during backpropagation, causing slow or no learning. Solved by ReLU activation, batch normalization, and residual connections (ResNet).

1

An Expert System (ES) mimics the decision-making of a human expert in a specific domain. Core components: Knowledge Base (domain facts and rules), Inference Engine (applies rules to facts), User Interface (interaction), and Explanation Facility (explains reasoning).

2

Forward Chaining (data-driven): starts with known facts, applies rules to derive new facts until the goal is reached. Used when all data is available upfront. Example: diagnostic systems.

3

Backward Chaining (goal-driven): starts with the goal and works backwards to find supporting facts/rules. Used when goal is known but evidence must be found. Example: Prolog, medical diagnosis with a specific disease hypothesis.

4

Knowledge Representation: facts and rules stored as IF-THEN productions. Example: IF temperature > 38 AND cough = yes THEN infection = probable. The inference engine pattern-matches these rules.

5

MYCIN is a classic expert system for diagnosing bacterial blood infections and recommending antibiotics. It was developed at Stanford and used certainty factors (degree of belief) instead of pure boolean logic.

6

Ant Colony Optimization (ACO) is a swarm intelligence algorithm inspired by ants' pheromone trails. Ants deposit pheromone on good paths; evaporation removes poor paths. Good paths accumulate more pheromone → positive feedback → convergence. Used for TSP, routing.

7

Particle Swarm Optimization (PSO) is inspired by bird flocking/fish schooling. Each particle has position and velocity. Particles update based on: pbest (personal best position) and gbest (global best position in swarm). Efficient for continuous optimization.

8

Fuzzy Logic handles imprecise or vague information. Unlike binary logic (true/false), fuzzy logic assigns degrees of membership between 0 and 1. Example: temperature is 'somewhat hot' with degree 0.7. Used in air conditioning control, washing machines.

9

Swarm Intelligence principles: decentralized control, self-organization, emergent behavior. No leader tells individuals what to do; complex global behavior emerges from simple local rules. Examples: ACO, PSO, Bee Algorithm.

10

Knowledge Engineering is the process of acquiring, organizing, and encoding knowledge from domain experts into a knowledge base. The knowledge engineer faces the knowledge acquisition bottleneck — experts find it hard to articulate tacit knowledge.