JU PMIT Exam Portal
Exam Countdown...
JU PMIT Exam Portal
Complete Exam Preparation Repository: Step-by-Step Numerical Math Solutions, In-Class Teacher Directives, Question Blueprint & Slide Decks.
Directives recorded in class notebook and re-verified via PMIT Batch-37 student discussions:
fillna(mean) / dropna(), column replacement, Tax calculations, conditional filtering).Complete step-by-step mathematical solutions to all recurring numerical questions from previous JU PMIT exam papers, along with comprehensive theoretical summaries:
NotebookLM only supports PDFs (PPTX is rejected). Download all 26 converted and original exam PDFs in a single zip:
Exam structure: 7 broad questions, answer any 5. Each question carries 12 marks (Total 60 Marks):
| Q # | Core Exam Topic | Question Details & Typical Problem | Question Type |
|---|---|---|---|
| Q1 | Computational Thinking & Logic | âĒ 4 CT Pillars (Decomposition, Pattern Recognition, Abstraction, Algorithms) âĒ Deductive vs Inductive reasoning (with examples) âĒ Imperative vs Declarative statements âĒ Converse, Inverse, Contrapositive truth rules |
Theory |
| Q2 | Time Complexity & Data Science | âĒ Big-O Analysis ($O(N)$, $O(N^2)$, $O(\log N)$, geometric series) âĒ Qualitative vs Quantitative data (Discrete vs Continuous) âĒ 1st, 2nd & 3rd party data differences âĒ Propositional logic with truth connectives |
Analysis |
| Q3 | Analytics Systems & Data Marts | âĒ 4 Analytics Types: Descriptive, Diagnostic, Predictive, Prescriptive âĒ PinkCity real-world energy optimization case study âĒ Correlation vs Regression analysis & applications âĒ Dependent vs Independent Data Marts âĒ 3 V's of Big Data (Volume, Velocity, Variety) |
Theory & Case |
| Q4 | KNN Classification [Clustering Excluded] |
âĒ KNN Numerical (High Priority): Euclidean distance step-by-step, ranking, majority vote with $K=3$ and $K=5$ âĒ Why KNN is a "Lazy Learner" (pros/cons, deferred computation) âĒ Note: K-Means Clustering is excluded per in-class directive |
â Numerical Math |
| Q5 | Linear Regression & Evaluation Metrics | âĒ Regression Line Formula: Slope ($b_1$), Intercept ($b_0$), predicting $\hat{y}$ âĒ Model Evaluation Metrics: MAE, MSE, RMSE, SST, and $R^2$ calculation âĒ Training vs Testing data split rationale |
â Numerical Math |
| Q6 | Statistics & ETL Architecture [ID3 & Neural Excluded] |
âĒ Variance & Standard Deviation: Mean $\bar{x}$, deviations $(x-\bar{x})^2$, sample variance ($s^2$) âĒ ETL Pipeline (Extract, Transform, Load) & Data Warehouse Metadata âĒ Note: ID3, Clustering, and Neural Networks are all excluded per in-class directive |
Math & Theory |
| Q7 | Python & Pandas (Mandatory Set) [1 Set Must] |
âĒ 1 Full Set Guaranteed: Handling missing values with fillna(mean) or dropna()âĒ Replacing values in columns (e.g. replace('Yangon', 'YG'))âĒ Computing Tax column ( df['Unit price'] * df['Quantity'] * 0.05)âĒ Conditional filtering & outlier clamping ( df.loc[df['Salary'] > 100000] = 75000)âĒ Matplotlib scatter plots & histograms |
â Python (Must) |
Topic-wise breakdown synthesized from peer analysis of previous semester finals, class tests (CTs), and in-class discussions, cross-referenced with exact lecture slides and teacher directives:
Definitions, Four pillars (Decomposition, Pattern Recognition, Abstraction, Algorithm Design), reasoning types (inductive/deductive), logical statements, and time complexity (O(1), O(N), O(N2)).
Data types (structured, semi-structured, unstructured), data marts, correlation vs. causation, regression analysis, and real-world commercial analytics case studies.
Logical expressions, truth tables, conditionals (p → q), biconditionals (p ↔ q), and logical transformations: converse (q → p), inverse (¬p → ¬q), contrapositive (¬q → ¬p).
Evaluation metrics: Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Total Sum of Squares (SST), Coefficient of Determination (R2), and step-by-step sample variance & standard deviation calculations.
K-Nearest Neighbors (KNN) classification, distance metrics, training vs. testing data split rationale, linear regression lines, with theoretical mentions of SVM and gradient descent.
The 3 V's (Volume, Velocity, Variety) and 5 V's (+ Veracity, Value), four categories of data analytics (Descriptive, Diagnostic, Predictive, Prescriptive), and the Big Data technology stack.
Data cleaning workflows, missing-value imputation (fillna, dropna), Pandas DataFrame operations, column replacements, conditional filtering, and Matplotlib visualizations.
The ETL pipeline (Extract, Transform, Load) and metadata classification into technical metadata, business metadata, and operational metadata.
A rigorous cross-reference against the 9-page past exam collection (PMIT_6122_CTDS_Past_Exam_Questions.pdf) and Dr. Rashed Mazumder's in-class directives revealed several high-probability subtopics omitted from the peer list, plus key scope boundaries:
The peer list mentions Propositional Logic, but past exams specifically test Predicate Logic (Slide Lec1-4): translating natural English statements into symbolic logic with universal quantifiers (∀) and existential quantifiers (∃), as well as negating quantified statements (¬∀x P(x) ≡ ∃x ¬P(x)).
Direct question from Exam Q1: contrast Imperative statements (step-by-step how to execute, e.g. procedural loops) against Declarative statements (specifying what result is wanted, e.g. SQL / logic programming).
Direct question from Exam Q2b (Slide 2-DA): 1st-party (direct customer interaction/CRM), 2nd-party (partner's 1st-party data shared via agreement), and 3rd-party (aggregated data acquired from commercial brokers).
Direct question from Exam Q3b: Dependent data marts draw cleansed data directly from a central Enterprise Data Warehouse (EDW), whereas Independent data marts source directly from operational systems without an EDW.
The peer list says generic "real-world case studies", but Slide 4-Predictive and past papers specifically examine the PinkCity Commercial Facility HVAC & Energy Optimization Case Study utilizing predictive regression and sensor telemetry.
Paired directly with the 12-mark KNN calculation in Exam Q4: explain why KNN is instance-based (no training phase; defers computation to query time) and discuss its tradeoffs (fast ingestion vs. high O(N × D) inference latency and memory storage).
The peer list includes SVM and Gradient Descent. Be aware that deep SVM hyperplanes and Gradient Descent optimization derivations have never appeared in JU CTDS exams. The instructor Dr. Rashed Mazumder explicitly clarified that advanced machine learning (Neural Networks, Clustering, SVM deep math) belongs to the AI course (PMIT-6107). Focus your study on guaranteed numerical problems (KNN, Linear Regression Line, Evaluation Metrics, Sample Variance, and Python Pandas).
Search and instantly view all official lecture presentation decks. Click "ðïļ View" to preview inline without force-downloading:
Official summer 2026 trimester final routine signed by Coordination Committee:
| Exam Date | Day | Time | Course Code & Title | Instructor | Venue | Status |
|---|---|---|---|---|---|---|
| 16 October, 2026 | Friday | 10:00 AM â 01:00 PM | PMIT-6122: Computational Thinking & Data Science | Dr. Rashed Mazumder (RM) | Room# 322 | Active Exam 1 |
| 23 October, 2026 | Friday | ð No Examination (Durga Puja Vacation Weekend) | ||||
| 30 October, 2026 | Friday | 10:00 AM â 01:00 PM | PMIT-6107: Artificial Intelligence & Neural Network | Prof. Dr. Mohammad Shahidul Islam (MSI) | Room# 322 | Exam 2 |
| 31 October, 2026 | Saturday | 02:00 PM â 05:00 PM | PMIT-6223: IOT and Fog Computing | Prof. Dr. Md. Sazzadur Rahman (MSR) | Room# 322 | Exam 3 |
| 06 November, 2026 | Friday | 10:00 AM â 01:00 PM | PMIT-6201: Cloud and Mobile Computing | Prof. Dr. Risala Tasin Khan (RTK) | Room# 322 | Exam 4 (Rescheduled) |