![](\"images/PyData.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# About Me\n",
"\n",
"
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"\n",
"# Delia Rusu\n",
"\n",
"\n",
"- Chief Data Scientist @ Knowsis\n",
"\n",
"- PhD, Natural Language Processing and Machine Learning\n",
"\n",
"- Interests: unstructured data & finance"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# This Talk\n",
"\n",
"![](\"images/ThisTalk.jpg\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# Financial Data\n",
"\n",
"![](\"images/FTSE100.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# Sources of Unstructured Data\n",
"\n",
"- Annual reports\n",
"\n",
"- Broker research\n",
"\n",
"- Conference call transcripts\n",
"\n",
"- Investor relations presentations\n",
"\n",
"- News and press releases\n",
"\n",
"- In-house content"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# What about unconventional datasets?"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# Wikipedia\n",
"\n",
"- 10 edits per second\n",
"\n",
"- the English Wikipedia \n",
" - currently has over 5M articles\n",
" - average of 800 new articles per day "
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"# Structured Views of Wikipedia\n",
"\n",
"- Wikipedia article -> vectorized representation of the context\n",
"\n",
"- Wikipedia as a graph"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"# Word2Vec\n",
"\n",
"- Word2Vec model (Mikolov et al. 2013)\n",
" - distributional hypothesis: words in similar contexts have similar meanings\n",
" - shallow, 2-layer neural network\n",
" - training objective - learn word vector representations which can predict nearby words (in context)\n",
" \n",
"![](\"images/SkipGramModel.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"# Word & Article Vectors\n",
"\n",
"- generate context vectors for:\n",
" - words which appear in Wikipedia articles\n",
" - Wikipedia articles themselves (topics)\n",
" \n",
"- model: wiki2vec\n",
" - https://github.com/idio/wiki2vec"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"source": [
"**Unilever** is an Anglo-Dutch **multinational consumer goods company**"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"source": [
"**Unilever_ID** is an Anglo-Dutch **multinational_consumer_goods_company_ID**"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"# Wikipedia Graph\n",
"\n",
"- Types of nodes:\n",
" - articles, categories\n",
" \n",
"- Types of directed edges:\n",
" - hyperlinks from article to article\n",
" - infobox links from article to article \n",
" - links from article to category\n",
" - links from category to category"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"# Graph-Based Similarity\n",
"\n",
"- using the distance between nodes in the graph\n",
" - in this case, $$Similarity(v_1,v_2) = 1 - Distance(v_1, v_2)$$\n",
" \n",
"- using random walks on graphs\n",
" - Personalized PageRank (Haveliwala, 2002) is a variation of PageRank (Brin and Page, 1998)\n",
" - the user query defines how important the vertex is, such that PageRank will prefer vertexes in the vicinity of the query vertex."
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# FTSE 100 Index Analysis"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": false,
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"data": {
"text/html": [
"| \n", " | Company | \n", "Ticker | \n", "Sector | \n", "Market cap | \n", "Employees | \n", "
|---|---|---|---|---|---|
| 0 | \n", "3i | \n", "III | \n", "Private equity | \n", "4.06 | \n", "266 | \n", "
| 1 | \n", "Aberdeen Asset Management | \n", "ADN | \n", "Fund management | \n", "3.14 | \n", "1,800 | \n", "
| 2 | \n", "Admiral Group | \n", "ADM | \n", "Insurance | \n", "4.91 | \n", "2,500 | \n", "
| 3 | \n", "Anglo American plc | \n", "AAL | \n", "Mining | \n", "6.09 | \n", "100,000 | \n", "
| 4 | \n", "Antofagasta PLC | \n", "ANTO | \n", "Mining | \n", "4.71 | \n", "4,005 | \n", "
| \n", " | Company1 | \n", "Company2 | \n", "W2V Similarity | \n", "
|---|---|---|---|
| 0 | \n", "3i | \n", "Aberdeen_Asset_Management | \n", "NaN | \n", "
| 1 | \n", "3i | \n", "Admiral_Group | \n", "NaN | \n", "
| 2 | \n", "3i | \n", "Anglo_American_plc | \n", "0.399286 | \n", "
| 3 | \n", "3i | \n", "Antofagasta_PLC | \n", "NaN | \n", "
| 4 | \n", "3i | \n", "ARM_Holdings | \n", "0.267260 | \n", "
| \n", " | Open | \n", "High | \n", "Low | \n", "Close | \n", "Volume | \n", "Adj Close | \n", "
|---|---|---|---|---|---|---|
| Date | \n", "\n", " | \n", " | \n", " | \n", " | \n", " | \n", " |
| 2013-01-01 | \n", "2909.5 | \n", "2909.50 | \n", "2909.50 | \n", "2909.5 | \n", "0 | \n", "2441.043 | \n", "
| 2013-01-02 | \n", "2930.0 | \n", "2969.30 | \n", "2923.00 | \n", "2969.0 | \n", "2394700 | \n", "2490.963 | \n", "
| 2013-01-03 | \n", "2969.5 | \n", "2982.50 | \n", "2952.50 | \n", "2980.0 | \n", "1763100 | \n", "2500.192 | \n", "
| 2013-01-04 | \n", "2992.5 | \n", "3005.00 | \n", "2975.21 | \n", "3000.0 | \n", "2041100 | \n", "2516.972 | \n", "
| 2013-01-07 | \n", "2998.5 | \n", "3004.99 | \n", "2976.00 | \n", "2985.0 | \n", "1185200 | \n", "2504.387 | \n", "
| \n", " | Company1 | \n", "Company2 | \n", "Correlation | \n", "
|---|---|---|---|
| 0 | \n", "3i | \n", "Aberdeen_Asset_Management | \n", "0.047557 | \n", "
| 1 | \n", "3i | \n", "Admiral_Group | \n", "0.067823 | \n", "
| 2 | \n", "3i | \n", "Anglo_American_plc | \n", "0.091343 | \n", "
| 3 | \n", "3i | \n", "Antofagasta | \n", "0.057116 | \n", "
| 4 | \n", "3i | \n", "ARM_Holdings | \n", "0.077546 | \n", "
"
]
}
],
"metadata": {
"celltoolbar": "Slideshow",
"kernelspec": {
"display_name": "Python 2",
"language": "python",
"name": "python2"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 2
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython2",
"version": "2.7.11"
}
},
"nbformat": 4,
"nbformat_minor": 0
}