Electricity Theft Detection ML Platform

$250-750 USD

Posted 1 minute ago

•

Ends in 23 hours

$250-750 USD

Paid on delivery

I need an end-to-end workflow for detecting Electricity non-technical losses, addressing class imbalance using SMOTE or cost-sensitive learning from AMI data. My raw information comes from smart meters and customer billing records, stored today as large CSV dumps that ultimately have to live in PostgreSQL. The signals that matter most are usage patterns, any evidence of meter tampering, and billing discrepancies. Here is the flow I have in mind. You start in Jupyter Notebooks, explore and clean the data, then build and compare Random Forest, SVM and a lightweight neural network so I can see which model produces the most reliable lift comparing accuracy, precision, recall, F1-score, confusion matrix, and ROC curves. Once the champion model is chosen, wrap it in a small Flask API for real-time scoring. A simple React dashboard should display the theft probability, key feature contributions and allow me to filter by account, date range or location. Please keep everything in Python (pandas, scikit-learn, TensorFlow/PyTorch—whichever you prefer) and make all queries database-agnostic so they run cleanly against PostgreSQL. The final repo should include: • Well-commented notebooks that reproduce the exploratory analysis, feature engineering and model training • The trained model artefacts plus a script to retrain with fresh data • Flask service with two endpoints: /score (single account) and /batch (CSV upload) • React front-end served separately or via the Flask app’s static folder • A README with setup commands and a one-line Docker build so I can spin the stack up locally Important: Clear documentation of the system is required. I’ll verify delivery by running the notebooks end-to-end on a sample dataset, checking that the API returns consistent probabilities, and confirming the dashboard renders correctly in Chrome.

Project ID: 40813071

About the project

8 proposals

Open for bidding

Remote project

Active 5 secs ago

Place your bid

Bid amount Email address

Benefits of bidding on Freelancer

Set your budget and timeframe

It's free to sign up and bid on jobs

8 freelancers are bidding on average $344 USD for this job

User Avatar
Flag of PAKISTAN

I can build the full detection workflow so your team can move from raw meter and billing CSVs to explainable theft-risk scores, then verify the results end to end. I’d load and validate the CSV data for PostgreSQL, then use reproducible notebooks to investigate usage patterns, tampering signals and billing discrepancies. To avoid overly optimistic results, I’d split data by time and account before preprocessing, apply SMOTE only to training data or compare cost-sensitive learning, and check for billing fields that could leak the answer. I’ll compare Random Forest, SVM and a lightweight neural network using your requested metrics and ROC curves, while giving particular attention to precision, recall and false-alert volume for this imbalanced problem. I’m comfortable building Python ML pipelines and connecting them to Flask and React. I’ll package the selected model with its preprocessing steps so notebook, /score and /batch probabilities stay consistent. The dashboard will show risk, feature contributions and account, date and location filters; database access will use PostgreSQL-compatible, portable queries. The repo will include commented notebooks, model artefacts, a retraining script, documented API examples, and a README with Docker build and local setup commands. I’ll check the supplied sample end to end and test the dashboard in Chrome. Do you have confirmed theft labels and a sample of the meter and billing CSV schemas?

User Avatar
Flag of INDONESIA

IF YOU ARE NOT HAPPY YOU DON'T PAY! The riskiest part here is the imbalance handling. If SMOTE runs before the split, or the same account's meter readings end up in both train and test, recall and ROC will look great in the notebook and then fall apart on real data. I'd split by account and time first, and apply SMOTE or class weights only inside the training folds of an imblearn pipeline. That same fitted pipeline (cleaning, features, model) is what gets saved and loaded by /score and /batch. That way the API returns exactly the notebook's probabilities, and the retrain script and React dashboard sit on top of it. Are the theft labels confirmed inspection results, or billing anomalies flagged by your team? Send me a message and I'll explain exactly how I'd approach your project.

User Avatar
Flag of PAKISTAN

Hello, I read "Electricity Theft Detection ML Platform" and the backend requirements are clear to me. How I would approach it: - Plan the data model, endpoints and authentication with you before writing code. - Build a secure, well documented API with proper validation, error handling and tests. - Deploy it, connect it to your frontend or third party services, and hand over with API documentation. I have built and maintained many backend systems and APIs, and I deliver clean, maintainable code. Could you share any existing documentation or the services this needs to connect to? Tarim

User Avatar
Flag of SOUTH AFRICA

Hi there, we might be new to the platform, but I’d love to share some relevant examples. Your project on detecting electricity non-technical losses sounds right up our alley, and we would love to exceed your expectations. I understand the importance of effectively addressing class imbalance with techniques like SMOTE or cost-sensitive learning, while ensuring a polished workflow from Jupyter Notebooks to a functional Flask API. Our approach will emphasize exploring usage patterns, meter tampering evidence, and billing discrepancies from your smart meter data. With multiple 5-star reviews on similar projects, we rank in the top 1% among 75 million users. While we might be new to Freelancer, we have over nine years of experience off-site. I truly appreciate you taking your time to review our proposal, and I would love to discuss this project more. Kind regards, Trichelle Client name is and my name is Trich123

User Avatar
Flag of GERMANY

Your smart meter and billing CSVs need to become a working theft scoring stack where tampering and billing gaps are flagged despite heavy class imbalance. I would load the dumps into PostgreSQL via SQLAlchemy so the queries stay database agnostic, then in Jupyter clean the data and build features from consumption patterns, tamper events and billed versus metered gaps. Random Forest, SVM and a small PyTorch network get compared with SMOTE against class weights, reporting precision, recall, F1, confusion matrices and ROC curves. The champion model goes behind Flask with /score and /batch, plus a React dashboard showing theft probability, feature contributions and filters for account, date and location, all started with one Docker build and documented in the README and notebooks. Based in Frankfurt, Germany, I have worked on data processing and machine learning in research since 2014 (PhD, 30+ publications). Do you have a labelled sample showing which accounts were confirmed theft cases, or should the labels come from inspection records in the billing data? Wahed

User Avatar
Flag of UNITED STATES

Hi there, Addressing class imbalance is crucial for your electricity theft detection model, especially since non-technical losses are rare events. Using SMOTE or cost-sensitive learning effectively will be key. I'll start by exploring your smart meter and billing data in Jupyter Notebooks, focusing on feature engineering to highlight tampering signals. We'll compare Random Forest, SVM, and a lightweight neural network, ensuring each model's results are interpretable with metrics like precision and recall. Once we select the best model, I'll wrap it in a Flask API for real-time scoring and create a React dashboard for insights. Best Andrew

内容来自 Freelancer News,版权归原作者所有。

阅读原文 ↗

把喜欢的内容,留在你的新标签页。

打开我的订阅