WBCHSE Applied AI study plan

Install a Python 3 toolchain (IDLE or VS Code plus Jupyter Notebook) and execute a script and a notebook cell successfully to confirm the environment works.

Write and debug expressions using the int, float, str and bool types, converting between them with int(), float() and str(), and predict each output before running the code.

Apply the four computational-thinking pillars (decomposition, pattern recognition, abstraction, algorithm design) to write pseudocode for a routine task such as splitting a restaurant bill with tip.

Define AI, machine learning, deep learning and rule-based systems in your own words, then classify five everyday products (spam filter, calculator, voice assistant, recommendation feed, chess engine) as rule-based or learning-based and justify each label.

Solve eight small problems using if/elif/else and while/for loops, including input validation and a running total, and comment what each block does.

Define and call your own functions with parameters and return values, writing at least three reusable helpers (for example a grade calculator and a temperature converter).

Choose between list, tuple, dictionary and set for four given scenarios and justify each choice in one sentence.

Implement a linear search and a simple bubble sort on a list and state the number of comparisons each performs on a sample of 10 items.

Distinguish supervised, unsupervised and reinforcement learning, giving one concrete example of each and naming the type of data each needs.

Load a CSV file into a pandas DataFrame with read_csv and inspect it using head(), shape, info() and describe() before changing anything.

Detect and handle missing values and duplicate rows using isnull().sum(), dropna(), fillna() and drop_duplicates(), stating the rule you applied and why.

Convert a column to the correct data type (for example text to datetime or numeric) and verify the change with dtypes.

Build four labelled charts with matplotlib (line, bar, histogram, scatter) including axis labels, a title and a legend where needed, and state in one sentence what each chart reveals.

Explain the difference between features and labels, and between training and test data, using your own dataset as the example.

Split a dataset into training and test sets with train_test_split, setting a fixed random_state and explaining why the split matters.

Train a scikit-learn classifier (for example a decision tree or logistic regression) on the training set and generate predictions on the test set.

Compute accuracy and a confusion matrix, then derive precision and recall by hand and describe when each metric matters more.

Diagnose overfitting and underfitting by comparing training and test accuracy, and describe two remedies such as more data or simpler models.

Summarise the full ML workflow - problem framing, data collection, cleaning, feature preparation, split, training, evaluation, deployment - as a numbered diagram.

Build a text preprocessing pipeline that lowercases, tokenises, removes stop words and applies stemming to a sample of 30 short review sentences.

Convert the cleaned text into numerical features using bag-of-words and TF-IDF, and explain in one paragraph how the two representations differ.

Train and evaluate a simple text classifier (for example Naive Bayes) on the labelled reviews and report accuracy on a held-out slice.

Explain how an image is stored as a grid of pixel values across RGB channels, and describe grayscale conversion, resizing and normalisation as preprocessing steps.

Describe conceptually how convolution and pooling layers let a CNN detect edges and then shapes, and name one documented bias risk in a real NLP or computer-vision system.

Map six AI application domains (healthcare, agriculture, education, finance, mobility, manufacturing) to one real deployed example each and state the data the system consumes.

Compare four AI tools (a chat-based large language model, Google Teachable Machine, TensorFlow Playground, a translation app) by input, output, cost and one clear limitation.

Explain the ethical principles of fairness, transparency, privacy, accountability and informed consent, giving a concrete violation of each.

Analyse one real AI deployment case, listing who benefits, who is harmed, which data was used and what safeguard was missing.

Distinguish emerging technology categories - generative AI, edge AI, AIoT and explainable AI - and state one realistic use and one risk for each.

Scope and complete a small AI project, writing a one-paragraph problem statement, the data source, the method chosen and the success measure before you start coding.

Package the project into a repository or portfolio folder containing the notebook, dataset or link, and a README covering problem, method, results and limitations.

Design a five-slide presentation (problem, data, method, results, next steps) and rehearse it aloud in under eight minutes.

Draft a one-page entrepreneurship idea applying AI to a local problem, naming the target customer, the value offered, the data required and a rough cost estimate.

Answer five likely viva or interview questions about your own project, including why you chose that model and what you would improve with more data.

Sit one timed 60-question mock covering all seven domains and tabulate your score per domain to expose weak areas.

Re-teach the five weakest concepts from the mock, then re-attempt at least 10 related questions and record the new score.

Re-run two practical tasks under time pressure (cleaning a CSV and training a classifier) without notes to confirm retention.

Compress all seven domains into a single two-page revision sheet of definitions, code snippets and evaluation metrics.

Rehearse the project viva for 10 minutes, answering follow-up questions on limitations and ethics.