Data Analytics F200 Paper D: Fundamentals of Data Analytics

A route number is a label with no order, so it is nominal. Ratings are ordered with uneven gaps, ordinal. Celsius has equal gaps but no true zero, interval. Time has a true zero, ratio.

Fixed columns are structured, JSON has tags but no fixed schema, and video and free text have no model at all.

JSON holds nested, labelled data without a fixed schema, which is why APIs use it.

Metadata.

A repeated record breaks uniqueness, a blank field completeness, an impossible date validity, and a value that cannot match reality accuracy.

Validation is an automatic check that data is reasonable and follows the rules: a range check would reject a journey time of 900 minutes on route 7, and a format check would reject 31/02/2026. Validation cannot tell whether sensible data is correct. Verification checks that data matches its source, for example by entering it twice and comparing, or by checking a record against the ticket machine's own log, so a mistyped fare of 2.50 instead of 2.00 would be caught.

Mean: 172 / 7 = 24.6. In order the times are 19, 22, 24, 25, 25, 26, 31, so the median, the middle value, is 25. The mode, the most common, is 25. The range is 31 - 19 = 12.

One journey held up by a breakdown or an accident could take an hour and pull the mean up, making every journey look slower. The median is the middle value, so a single extreme journey barely moves it.

54,000 - 48,000 = 6,000. 6,000 / 48,000 = 0.125. 0.125 x 100 = 12.5%.

Correlation does not show causation. Both figures are probably driven by a third factor, such as tourists in the summer, who both take more buses and buy more from the café.

Extract takes the data out, transform cleans and reshapes it, and load writes it to the destination.

A data lake stores everything raw, with schema on read. A warehouse applies its schema when data is written.

A transport the systems cannot share is technical, a structure that does not match is syntactic, and a different meaning is semantic.

A second factor is MFA, minimum access is least privilege, a record of access is audit logging, and permissions by role is RBAC.

Cloud storage can be scaled up in minutes when the company starts collecting GPS data from every bus, and scaled down again, so it pays only for what it uses instead of buying servers up front. The provider also maintains and patches the platform and keeps copies in several data centres, so the data survives a hardware failure without the company employing its own specialists.

The seven principles are lawfulness, fairness and transparency; purpose limitation; data minimisation; accuracy; storage limitation; integrity and confidentiality; and accountability.

A copy is access, a correction is rectification, deletion is erasure, and a machine-readable copy to move elsewhere is portability.

Publish totals for each stop rather than individual journeys, with smartcard numbers removed and times grouped into bands such as hours. Counts below a minimum, say ten, should be suppressed, because a quiet stop used by one regular passenger could otherwise reveal when that person travels.

Engineers build pipelines, scientists build models, analysts report, and the data protection officer oversees compliance.

A line chart shows change over time.

The company should first define the question precisely: which routes, days and times are over or under capacity, and where journeys run late. It can then combine the three sources, ticket sales for demand, GPS for actual journey times and surveys for passengers' views, after agreeing definitions so they join correctly, for example whether a return ticket counts as one journey or two. The data must be cleaned before analysis: duplicates removed, dates standardised and missing fares handled. Outliers, such as a 900 minute journey, should be investigated rather than deleted, because a real breakdown pattern matters for a timetable. The cleaned data should be loaded into the warehouse, with access limited to the staff who need it. Analysis should compare demand by route, day and hour, using medians so that occasional delays do not distort typical journey times, and trends across months to allow for seasonal tourists. The findings should be presented to managers in plain language with a few clear charts, such as a line chart of demand through the day, and specific recommendations, such as an extra bus on route 7 at 8am. Legally, smartcard journeys are personal data under the Data Protection Act 2018 and UK GDPR. The company needs a lawful basis, and using ticketing data for planning must be compatible with the purpose it was collected for. It should minimise the data, pseudonymise or aggregate it before analysis, keep it secure and delete it when it is no longer needed, and survey respondents must have been told how their answers would be used. Overall, careful preparation and aggregation let the company plan timetables from real demand while protecting passengers' privacy.