Campus Recycling Collection with Machine Learning: A 2027 Project Guide

A convincing Machine Learning portfolio project explains the result as well as showing it. Here, a environment-club volunteer asks: How much verified material weight was collected by type? You will build a baseline and evaluation experiment and follow the relevant records from a small fixture to an independently checked answer.

This is a 2027 study-edition project brief. All example records and operational situations are fictional. Start with basic Machine Learning knowledge, the ability to inspect structured data and a local practice environment. The complete first version should answer the stated question before adding accounts, real integrations or additional screens.

Define the project rule

Verified collected weight uses one unit and excludes rejected records. The result records a practice collection total, not carbon savings or an environmental certification. For this exercise, one row represents one collection. The stable key is record_id; material_type is the grouping or filtering label. The numeric weight_kg field uses kilograms. Accepted states are planned, verified, rejected.

The summary selects status verified and applies sum to the selected numeric values. Keep this definition visible in the interface, response or report so another learner can reproduce it. Use positive finite values in the starter fixture and define any future zero or missing-value policy explicitly.

Choose the Machine Learning approach

Choose a target only if it serves a real question and is available for an adequate set of labelled examples. Build a preprocessing-and-model pipeline, compare it with a simple reference and examine failure cases. Keep a held-out evaluation set. The supplied fixture is for schema tests; obtain a suitable synthetic training dataset before fitting a model.

The three design decisions to explain in your project notes are:

• Target validity. Apply this choice to collection records and explain its effect on weight_kg.
• Preprocessing boundaries. Apply this choice to collection records and explain its effect on weight_kg.
• Honest held-out evaluation. Apply this choice to collection records and explain its effect on weight_kg.

Work through a small sample

A separate predictive extension needs an adequately sized labelled synthetic dataset, a declared target and held-out evaluation. The example below only checks the processing contract.

[
  {
    "record_id": "recycling-collection-01",
    "material_type": "Paper",
    "weight_kg": 2.4,
    "status": "verified",
    "recorded_on": "2027-01-04"
  },
  {
    "record_id": "recycling-collection-02",
    "material_type": "Cardboard",
    "weight_kg": 5.1,
    "status": "planned",
    "recorded_on": "2027-01-05"
  },
  {
    "record_id": "recycling-collection-03",
    "material_type": "Plastic",
    "weight_kg": 1.6,
    "status": "verified",
    "recorded_on": "2027-01-06"
  }
]

Only the first and third records match verified. Their weight_kg values are 2.4 and 1.6; the full-fixture result is 4 kilograms, with 2 matching records. Selecting material_type = Paper leaves the first record, so the result becomes 2.4 kilograms, with 1 match. This provides two independent checks for the filter and aggregation.

The full-fixture output contract is:

{
  "matching_count": 2,
  "operation": "sum",
  "value": 4.0,
  "unit": "kilograms"
}

Check the result by material_type

The overall figure answers the all-records version of the question. Grouping by material_type produces the following separate expectations, using the same verified rule. The middle group has a record in the source fixture but no qualifying observation.

Use this breakdown when answering the group-level question. Compare like-for-like units and keep the matching count visible, especially for averages. A different filter changes which records enter the calculation; it does not change the meaning of the metric.

The code blocks describe sample data and an expected result. Implement the surrounding feature using the installed version of your chosen tools, then save the actual run evidence separately from these expectations.

Build the feature in three passes

1. Establish a non-predictive baseline

First answer the descriptive question using the sum of weight_kg for verified records. This creates a verified baseline and exposes schema problems. Three sample collection records support processing tests, not a meaningful training experiment.

2. Define a defensible extension

Propose a synthetic target such as whether a future weight_kg exceeds a threshold chosen before training. State when the prediction would be made and which material_type and status information would be available then. Obtain enough labelled synthetic examples to study the question.

3. Evaluate without leaking the answer

Separate training and evaluation data before fitting preprocessing. Compare the model with a simple reference and inspect errors by material_type. Report the data construction and limitations. Keep measured evaluation results out of the article until the experiment has actually run.

Find the mistake before polishing the result

A useful defect to introduce deliberately is: An unverified estimate is included as measured collection weight. First reproduce it against the fixture, then identify the smallest rule or state transition that explains the mismatch. Fix that boundary and rerun the full-fixture and selected-label examples. This creates a stronger explanation than changing several unrelated parts until the screen looks right.

The scenario-specific acceptance check is: Keep estimates separate and add weight only after verification. Preserve before-and-after evidence. Add a missing material_type, a negative weight_kg, an unrecognised status and a duplicate record_id as separate invalid cases. Define the rejection or duplicate-handling policy before testing, and verify that invalid input does not silently change valid existing records.

Check the Machine Learning concerns as well:

• A preprocessing step must not learn from the test set.
• Class imbalance or target spread must be reported.
• No model score should be inferred from the tiny fixture.

For an empty filter match, the expected sum value is 0. Keep matching_count at zero. This distinction prevents an empty dataset from being interpreted as a measured observation.

Explain the finished work in a portfolio

Show the environment-club volunteer question, the record definition and the expected calculation first. Then demonstrate a material_type filter and the acceptance check. Include the fixture, setup instructions, chosen tool versions, actual test output and one remaining limitation in the repository. If you show a screenshot, pair it with the input that produced it. Another learner should be able to repeat the same result from a clean checkout.

For a focused extension, add a second material_type selection and compare the two groups using the same rule. Specify how missing groups appear and keep the original fixture as a regression case. Expand scope only after the current calculation, state handling and recovery path remain understandable.

Questions learners often ask

Which collection records belong in the result?

Select records whose status is exactly verified and that satisfy the chosen material_type filter. Other states remain in the source data and can have separate summaries. Follow the domain rule above when a record changes state.

What should happen when no records match?

Return a zero total with a matching count of zero and a clear empty-result message. Keep the applied filter visible so the user can change it. Avoid using a loading indicator for a completed request that simply has no matching data.

How can I prove the weight_kg result is correct?

Reconcile the first and third sample records, then repeat the calculation for Paper alone. Save both expected and actual outputs, and rerun the scenario-specific acceptance check after making a change.

Continue learning with Groot Academy

Use this exercise as a concrete discussion point when choosing a learning path. Explore the related Groot Academy course and ask about the current syllabus, learning mode and project practice. Course availability and batch details should be confirmed directly with the academy.

For implementation details, consult the official Machine Learning documentation matching the tools you install. This project defines an original practice brief; it does not depend on a forecast about 2027 software releases.

Continue with Water Refill Station Checks using Machine Learning, or compare Campus Recycling Collection using Microsoft Excel. These companion practice guides are part of the 2027 learning collection.