SensexPrediction

A proof-of-concept M.Tech project for predicting Sensex stock trends using NLP and SVM, built in C# with Visual Studio 2010.

Overview

• Purpose: Analyzes news sentiment and technical indicators (EMA) to predict Sensex stock movements.
• Tech: C# (.NET 4.8), OpenNLP (NLP), LibSVM (SVM), SQL Server (data storage), HtmlAgilityPack (web scraping), Twitterizer (Twitter integration).
• Status: Academic prototype; not suitable for real-money investing.

Directory Structure

• Assemblies/: External DLLs (e.g., HtmlAgilityPack.dll, OpenNLP.dll).
• database/: StockPredictData_Schema.sql for SQL Server schema.
• Models/: OpenNLP models (EnglishSD.nbin, EnglishTok.nbin, EnglishPOS.nbin, Parser\tagdict).
• Properties/: Project metadata and settings.
• Resources/: UI images.
• Root: Source .cs files, .csproj, .sln, thesis (MTech Thesis.pdf).

Setup

1. Prerequisites:
   • Visual Studio 2010+ (with .NET 4.8).
   • SQL Server (e.g., Express edition) for the StockPredictData database.
2. Steps:
   • Copy external DLLs (HtmlAgilityPack.dll, Newtonsoft.Json.dll, OpenNLP.dll, etc.) to Assemblies/.
   • Update app.config with your SQL Server instance (e.g., Server=NEURALLAP\SQLEXPRESS).
   • Open SensexPrediction.sln in Visual Studio.
   • Run database/StockPredictData_Schema.sql in SQL Server Management Studio to create the database.
   • Build the solution to generate SensexPrediction.exe in bin\Debug\.
   • Run SensexPrediction.exe.

Dependencies

• External DLLs (in Assemblies/):
  • HtmlAgilityPack.dll: Web scraping.
  • Newtonsoft.Json.dll: JSON parsing (used by Twitterizer).
  • OpenNLP.dll, SharpEntropy.dll, SharpWordNet.dll: NLP processing.
  • SVM.dll: SVM classification.
  • Twitterizer.dll: Twitter API integration.
  • Microsoft.VisualBasic.PowerPacks.Vs.dll: UI controls (optional, check Sensex_Prediction_Form.cs).
• NLP Models: Included in Models/ (from OpenNLP).
• Note: Update [Miscellaneous].ModelPath in the database if the Models/ path changes.

Model Generation

• NLP: SharpNLP.cs uses OpenNLP models from Models\ for feature extraction.
• SVM: SVM.cs and SVMMapping.cs handle model training and mapping, saved to [Miscellaneous].ModelPath (e.g., svm_model.txt). History stored in [SVMModel] table; no filesystem backups confirmed.
• Paths: Hardcoded paths updated to relative (e.g., Models\, Dump.txt).

Thesis

• MTech Thesis.pdf: Full documentation of methodology and results.

Limitations

• Relies on deprecated Google Finance URLs.
• Twitter integration uses Twitterizer 2.3.2 (pre-API v2).
• Proof-of-concept only; lacks real-time reliability.

Notes

• Helper functions are available under Miscellaneous folder.

#AI #MachineLearning #Finance #Sensex #NLP #Csharp