Team members:

• Anuja Jadhav
• Komal Mehta
• Padma Narayanaswamy

Data:

• HTTP REST API - “NEWS API”
• Secondary source - Marketstack API for analyzing our stock prediction
• Data cleaning and transformation on the resulting JSON response.

Data collection -

• Custom code for fetching the data from NEWSAPI

• Tech Giants - Microsoft, Facebook, Twitter, Apple, Meta, Tesla, Intel, Alphabet, IBM, NVIDIA, Netflix, Salesforce

• Data collated into a CSV file and used for further processing

Problem Description:

• Perform comprehensive analysis on NewsApi using Natural Language Processing (NLP).

• Helps us in analyzing the emotions and predicting stock movements.

• Helps us in mitigating the risk involved and make informed decision whether to hold or sell stocks.

• Market hypothesis assumes that a stock’s performance would have its impact from historical information or current events.

• With the advent of Artificial Intelligence and Machine learning algorithms, we could make predictions in the closest range possible.

• This inspired us to pursue this analysis to learn more about market efficiency.

Analytics plan:

• Preliminary data analysis also called as Exploratory data analysis to get the overview of shape of the data and get more insights

• Based on the results and further analysis we will perform data cleaning such as removing invalid or absurd values/ stopwords and transform the data

*For the same we will be using tools such as RStudio, RMarkdown and libraries such as keras, ggplot2, tm etc. using methods such as Natural Language Processing, Sentiment Analysis .

Word Cloud

Count of Sentiments

Evaluation Plan:

• Evaluate our results based on the sentiment and corresponding stock movement.
• Showcase this by ggplot2 visualization, tm for textmining and Keras that’s holding the fort!

Responses Summary

• Regarding Other API - we have used market watch

• Considering keywords for analysis

• Data Cleaning involved clearing extra symbols and stopwords.

• Reason for Keras : Keras is easy to use, implementing neural network is very easy in this, with just one line of code we can add one layer in the neural network with all it’s configurations. It comes with lot of data processing libraries. We are factorizing the sentiment which also makes your development easy and fast.

Procedure Summary

• The model states that tech firms had more positive news in the past 30 days
• Also we arrived at 72% accuracy which added more weight to our prognosis
• To confirm our analysis further we compared the stock movement of Microsoft using marketstock API and google trends to track number of hits happened for stock MSFT -Microsoft

Loss Vs Accuracy Plot

Number of Hits - Google Trends for Microsoft

Microsoft Stock Movement

Result Summary

• Solid reliable outcome considering 12 firms and restricted dataset provided by the NewsAPI

• Microsoft stock movement for over a period of time, clearly shows increase in the value in the recent 30 days of time.

• W can also include intraday movement to better understand the stock movement followed by a specific event such as Q-results.

• This will help the investors to take better informed decision in retaining or selling the stocks.

Key take-aways

• Considering the effective accuracy of the model for our dataset with the limited time period (1 month data) proves that it could be further improved if the dataset is diverse for a longer period of time.

• Same technique can also be applied to other domains such as Sports, Banking, Healthcare, etc.

• More analysis and prediction could be performed using historical data of various firms.