Fundamentals of Business Intelligence & Data Science

The Excel practice for this subject lives at bids.goatedd.tech, not here. This portal carries the theory: what each method is, when to use it, how the numbers are computed by hand on a small example, and how to read an output. Every chapter's recap points back at the lab.

Material is adequate for the pre-mid half and thin on prose for the post-mid. Section A has three slide decks (32 pages in total), one 1,292-word notes file, eight Excel workbooks, and eight complete R session documents. Two textbooks are in the Drive: Camm et al. Business Analytics 5e (1,037 PDF pages) and Provost and Fawcett, Data Science for Business (409 pages, background only). Every textbook citation here is to Camm 5e, because the outline's primary text, An Introduction to Management Science, and the Albright and Winston reference are both absent from the Drive.

The single biggest gap. POST MID SEM/BUSINESS INTELLIGENCE/SLIDES NOTES/BIDS NOTES.pdf is 51 pages, scanned, with no text layer. It is almost certainly the handwritten post-mid notes and could not be read. Open it yourself, especially before the simulation and forecasting chapter, which has no other source.

Chapters built largely from the textbook and general knowledge, each labelled "Check against your class slides." where it applies: the whole of Monte Carlo simulation and forecasting new products (no slides, no notes, no class file in either section, and the Bass model is not named anywhere in the material); the whole of Bayes' theorem and Bayesian networks (the 17 August announcement shipped only a dataset and an instruction to install Orange); everything in non-linear optimisation beyond the Solver engine names, since the WSJ workbook carries a Solver payload and no explanatory text; and how Orange itself is driven.

Two errors in the source material, corrected in place and flagged where they appear. The Confusion Matrix deck's specificity slide says "10 actual non-buyers, correctly predicted 8, so 80 %", which contradicts the same deck's own accuracy, precision and recall figures; those fix the matrix at TP 6, FN 2, FP 4, TN 8, so there are 12 actual non-buyers and specificity is 8/12 = 66.67 %. And the collaborative-filtering workbook's denominator cell is an array formula that could not be read: summing the correlations signed gives a prediction above 22 on a five-point scale, so the chapter teaches the standard absolute-value form, which gives 3.66.

Section B only: the session-1 deck (FBIDS Class-1 PPT.pdf, 27 pages) is the only slide source for the opening chapter, and supplies the DIKW ladder, the datafication framing, the quantitative/qualitative data tree on a real AQI table, and the 4 Vs. Section A's Classroom has no session-1 deck at all.

Section B only: the LP deck and its live workbook (Linear Optimization Models.pdf, LP Excel work (for Class).xlsx) teach one tea-stall problem three ways, formulation then graph then Solver, and add a self-contained binary hospital-equipment problem Section A does not have.

Section B only: the ML Concepts deck (24 pages) is the strongest single source on classifier evaluation in either section, adding the cutoff sweep at several probabilities, the ROC curve and AUC, and the five-step data-mining process with its 80/20 split. Its companion Performance Metrics Calculation.xlsx builds the confusion matrix from four COUNTIFS and drives the predicted class from a changeable cutoff cell.

Section B only: the clustering deck and its solution workbook add Euclidean distance named as the standard measure for continuous data, the WCSS elbow for choosing k, and one full k-means iteration visible in Excel formulas.

Section B only: the decision tree material (DT Solution excel (1).xlsx) is an entire topic Section A never taught, so it has its own chapter, titled and labelled "Section B only". It computes root entropy, conditional entropy per branch and the weighted entropy of each candidate split in Excel, and concludes on the sheet that Income has the lowest weighted entropy and is chosen first.

Section B only: multicollinearity and VIF appear nowhere in Section A's material. MLR Solution Data (1).xlsx computes R squared 0.964 for wait time, a VIF near 28, and annotates it "Remove this Variable", then refits: age moves from p = 0.90 to p = 0.013. That whole episode is built into the regression chapter.

Section B only: the PCA deck is the only PCA slide source in either section, supplying the meta-variable definition, the correlation matrix that motivates reduction, the six-variable eigenvalue table, and two rules for choosing the number of components (90 % cumulative variance and the Kaiser criterion).

Section B only: the group assignment, a 6 x 5 movie ratings matrix with a genre column and eight cells to predict, is the applied exercise for the recommender chapter and the source of its content-based worked example.

Section B only: the assessment detail that the Quiz component is two quizzes of 10 marks each, that Individual Assignment and PPT is marked NA, and that Power BI was required alongside Excel and R or Python.

Two Section B files are scanned with no text layer and could not be used: knn note (1).pdf (2 pages) and decision tree note (1).pdf (4 pages). Both are another section's handwritten notes; check them against your class slides if you can get them.

Section B adds nothing on market basket analysis, Bayesian networks, R itself, Monte Carlo simulation or forecasting, so those chapters rest on Section A's Classroom files and the student's own Drive notes alone. Section B confirms the same outline, the same credits, the same weights and the same reference books, and does not change the fact that this exam is on a computer. Note also the spelling: Section B's roster gives Maji where this course's outline writes Majhi.