<p>Have tried to create a one-stop machine learning solution for financial use cases.&nbsp;</p><p>It offers:</p><ul><li>Sentiment analysis on financial headlines/tweets</li><li>Stocks forecasting using time series analysis.&nbsp;</li></ul><p>Kindly give it a try:</p><p>Coming soon </p><ul><li>Training pipeline for classification task</li><li>Inference functionality for NER, text generation </li></ul>

Fin-ML

<h2>Quickstart</h2><p>This model is trained on financial headlines dataset. I have used "small_bert_en_uncased" for training the dataset.</p><PRE><code>from Fin_ML import fin_nlp, stocks_forecast
loaded_model = fin_nlp.load_classifier_model()

# Input is called inside the function, 
prediction = fin_nlp.predict_single_sentiment(model)

Eg : Investors beware — vaccines aren’t a silver bullet for markets
prediction
prints -- ('negative', 0.9195326)

# Stocks forecasting using Exponential Smoothing ("stock", "days to forecast", "Values - ['Open', 'High', 'Low', 'Close', 'Adj Close', 'Volume']")
stocks = stocks_forecast.exp_smoothing_forecast("TSLA", 8, "Open")

# Stocks forecasting using Prophet ("stock", "days to forecast", "Values - ['Open', 'High', 'Low', 'Close', 'Adj Close', 'Volume']")
stocks = stocks_forecast.prophet_forecast("TSLA", 8, "Open")
</code></PRE><h3>Benchmark with other sentiment tools</h3><p>I have taken a sample of 100 sentences involving financial terms and analyzed it with the various sentiment analysis tools in the market and below are the observations. Due to class imbalance in the dataset I have chose F1-score for benchmarking the tools</p><table dir="ltr">
<thead dir="ltr">
<tr dir="ltr">
<th dir="ltr">Sentiment analysis tool</th>
<th dir="ltr">time take for inference</th>
<th dir="ltr">F1-score</th>
</tr>
</thead>
<tbody dir="ltr">
<tr dir="ltr">
<td dir="ltr">Fin-ML</td>
<td dir="ltr">363 ms</td>
<td dir="ltr">0.82</td>
</tr>
<tr dir="ltr">
<td dir="ltr">Vader-sentiment</td>
<td dir="ltr">17.6 ms</td>
<td dir="ltr">0.26</td>
</tr>
<tr dir="ltr">
<td dir="ltr">transformers-pipeline</td>
<td dir="ltr">4910 ms</td>
<td dir="ltr">0.30</td>
</tr>
</tbody>
</table><p>P.S : I will be adding more tools for benchmarking</p><h2>Coming soon</h2><UL><LI>Training functionality for classification models</LI><LI>Exploring on NER, text generation and will be adding soon in the future</LI></UL>

Deep learning applications on the financial and banking world - GitHub - Vishnunkumar/FinML: Deep learning applications on the financial and banking world

moneyandinvestments

Money & Investments

artificialintelligence

Artificial Intelligence

<p>This model is trained on financial headlines dataset. I have used "small_bert_en_uncased" for training the dataset. </p>

<p>I have taken a sample of 100 sentences involving financial terms and analyzed it with the various sentiment analysis tools in the market and below are the observations. Due to class imbalance in the dataset I have chose F1-score for benchmarking the tools </p>

<p>P.S : I will be adding more tools for benchmarking </p>

<ul><li>Training functionality for classification models </li><li>Exploring on NER, text generation and will be adding soon in the future </li> </ul>

GitHub - Vishnunkumar/FinML: Deep learning applications on the financial and banking world

Fin-ML

Vishnu Nandakumar's ideas are part of this journey:

Related collections

Similar ideas