Kaggle_submissions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Creating files for submission to Kaggle  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "import csv\n",
    "import pickle\n",
    "# internal imports\n",
    "\n",
    "import helpers as HL\n",
    "import glove_module as GV\n",
    "import validation_and_prediction as VP\n",
    "import neural_nets as NN\n",
    "import tokenizing as TO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Making corpus:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "full_corpus, nr_pos_tweets, nr_neg_tweets, total_training_tweets = HL.get_corpus(full=False, test=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create the word embeddings using the created gensim-.txt file."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Pick one, the higher dimension, the better result and longer computational time. \n",
    "global_vectors=HL.get_global_vectors(200)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing unprepocessed full dataset: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_name=\"keggle_glove_unprepro_simple.csv\"\n",
    "pred= VP.get_prediction(NN.basic_model_adam, global_vectors, full_corpus, total_training_tweets, nr_pos_tweets,kaggle_name, patience=10, epochs=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing best preprocessing with full dataset: \n",
    "## Applying the preorocessing:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "##Create on your own:\n",
    "\n",
    "input_=[{'hashtag': True, 'segmentation_hash': True, 'hashtag_mention':True,\n",
    "        'hearts':True,'hugs_and_kisses':True,'elongation':True, 'set_to_not':True}]\n",
    "final_corpus=TO.preprocess_corpus(full_corpus, input_)\n",
    "final_corpus_ngram=HL.creating_n_grams_corpus(2,final_corpus)\n",
    "stopwords=TO.get_dynamic_stopwords(best_corpus, MinDf=100, MaxDf=0.8)\n",
    "final_corpus_ngram_stopwords=TO.remove_stopwords(final_corpus_ngram, stopwords)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## or upload from pickle\n",
    "final_corpus_ngram_stopwords=pickle.load( open( \"stopword100_corpus_n2_SHM_E_SN_H_HK.pkl\", \"rb\" ) )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Making kaggle submission using simple neural net: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_name=\"keggle_glove_best_simple.csv\"\n",
    "pred= VP.get_prediction(NN.basic_model_adam, global_vectors, final_corpus_ngram_stopwords, total_training_tweets, nr_pos_tweets,kaggle_name, patience=10,epochs=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Making kaggle submission using complex neural net: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "kaggle_name=\"keggle_glove_best_complex.csv\"\n",
    "#final_corpus=n_grams_corpus\n",
    "pred= VP.get_prediction(NN.complex_model, global_vectors, final_corpus_ngram_stopwords, total_training_tweets, nr_pos_tweets,kaggle_name, patience=10,epochs=100)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}