Video Search & Recommendation

Demo Preview

App & User Interface

The entry point is App.java, a JavaFX Application that loads the FXML scene Searching.fxml into a 1920×1080 stage:contentReference[oaicite:0]{index=0}.

• FXML: Defines UI layout with a search bar, scroll pane for results, and toggle buttons.
• Controller: PrimaryController.java handles search button events, toggling search mode (title vs tags), and passing queries to the algorithm:contentReference[oaicite:1]{index=1}.

Vdeo Data & Nodes

Each video is wrapped in a MarkovNode containing:

• Video ID + metadata (title, tags, views, likes, dislikes, comments):contentReference[oaicite:2]{index=2}
• Arrays of linked node IDs (ProbIDs) and their probability coefficients (ProbCoefficients).

Nodes track how likely one video leads to another. Clicking videos updates probabilities, strengthening links between them:contentReference[oaicite:3]{index=3}.

Markov Chain Model

The MarkovChain class builds a probabilistic network of video nodes:

• On initialization, every video becomes a node:contentReference[oaicite:4]{index=4}.
• Each pair of videos gets an initial probability coefficient.
• User interactions call alterProbability(), reinforcing the edge from a previous node to the clicked node:contentReference[oaicite:5]{index=5}.

User clicks video B after video A:
→ prevNode.changeProb(B, +1)
→ Increases probability coefficient
→ Video B more likely to appear in future searches
      

Search Algorithm

The SearchAlgorithm class implements fuzzy matching + weighted ranking:

• Uses FuzzyWuzzy (partialRatio) to match search term against titles or tags:contentReference[oaicite:6]{index=6}.
• Scores videos with a weighted formula:
  • Comments × 15
  • Likes × 10
  • Views × 1
  • Dislikes × –5:contentReference[oaicite:7]{index=7}
• Normalizes scores relative to max probability.
• If a previous click exists, blends click probabilities with content-based relevance:contentReference[oaicite:8]{index=8}.
• Sorts results numerically for display:contentReference[oaicite:9]{index=9}.

Result: A personalized, evolving ranking of videos.

Controller Flow

The UI’s PrimaryController ties everything together:

1. User types a query in txtSearch.
2. Clicks Search → calls SearchAlgorithm with current MarkovChain:contentReference[oaicite:10]{index=10}.
3. Receives a Hashtable<MarkovNode, Double> of results (video nodes with probabilities).
4. Passes results to generateImaging() for rendering thumbnails + stats.
5. User clicks a video → updates Markov chain, reinforcing personalization:contentReference[oaicite:11]{index=11}.

Summary of Systems

• UI Layer: JavaFX FXML + Controller
• Data Layer: Video objects wrapped in MarkovNodes
• AI/Logic: MarkovChain + SearchAlgorithm
• Feedback Loop: User clicks → adjust probabilities → improve personalization