TradeInsight

An AI demo model predicting market moves, built on coffee and code

Comment

Inspiration : The idea for TradeInsight came from my own early trading experiences—where I often wished I had something smarter than just gut feeling. Back then, AI wasn’t as accessible, and looking back, I can’t help but think: if I had AI, maybe I wouldn’t have lost money so quickly. Fast forward to today, with AI becoming mainstream, I decided to ask: what happens when we mix AI with trading? That question sparked the start of this project.

What it does :

  1. TradeInsight is a demo model that takes historical trading data and attempts to predict market trends.
  2. Processes datasets with Open, Close, High, Low, and Volume.
  3. Uses machine learning models to generate predictions.
  4. Visualizes actual vs. predicted values through a clean dashboard.
  5. Currently built as a prototype for a hackathon project.

How we built it :

  1. Data Collection & Cleaning: Collected sample market datasets and normalized them.
  2. Model Training: Implemented ML algorithms (starting with Linear Regression and expanding into others).
  3. Visualization: Used Python libraries (Matplotlib, Seaborn) for performance graphs.
  4. Dashboard: Built a simple interface to present predictions in a user-friendly way.

Challenges we ran into :

  1. Handling noisy and incomplete financial data.
  2. Achieving accuracy beyond random guessing.
  3. Designing a dashboard under time pressure (navbar alignment alone took hours).
  4. Managing the balance between hackathon deadlines and learning advanced concepts.

Accomplishments that we're proud of :

  1. Built a working AI-driven prediction model from scratch.
  2. Completed a functional demo dashboard despite UI hurdles.
  3. Learned how to integrate data science with product design.
  4. Pushed through late nights, deadlines, and coffee-fueled coding sessions to deliver something meaningful.

What we learned :

  1. The importance of data preprocessing in financial ML.
  2. The math behind loss functions and optimization: 𝐽(𝜃)=12𝑚∑𝑖=1𝑚(ℎ𝜃(𝑥(𝑖))−𝑦(𝑖))2J(θ)=2m1 ​i=1∑m (hθ (x(i))−y(i))2
  3. That building models is easier than deploying them in real-world apps.
  4. UI/UX matters as much as predictive accuracy.

What's next for TradeInsight :

  1. Connect APIs for real-time trading data.
  2. Explore advanced ML models (Random Forest, LSTMs for time-series).
  3. Improve accuracy through feature engineering.
  4. Deploy the dashboard as a web app for easier access.

Built With

Share this project:

Updates