File size: 1,331 Bytes
0063cea
 
 
dfedb39
 
 
 
0063cea
023ee95
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
dfedb39
 
 
023ee95
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0063cea
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
---
license: mit
tags:
  - time-series
  - forecasting
  - stock-prediction
library_name: scikit-learn
---
#  DataSynthis_ML_JobTask

##  Task Overview
This project focuses on **Time-Series Forecasting of Stock Prices**.  
We used historical stock data to forecast future closing prices.

##  Models Implemented
- **ARIMA** (Traditional Statistical Model)  
- **LSTM** (Deep Learning Model)  
- **Prophet** (Optional – if used)  

##  Dataset
- Public stock dataset from [Yahoo Finance](https://finance.yahoo.com/).  
- Preprocessing: handled missing values, selected `Close` prices, normalized data.  

##  Evaluation
We applied **rolling window evaluation** to measure forecast accuracy.  

### Performance Comparison

| Model    | RMSE   | MAPE   |
|----------|--------|--------|
| ARIMA    | 14.23   | 5.92%  |
| LSTM     | 9.87   | 4.35%  |
| Prophet  | 11.45   | 5.10%  |

##  Results & Recommendation
- **LSTM** generalized better, capturing long-term patterns.  
- **ARIMA** worked for short-term stable data.  
- **Prophet** was useful for trend/seasonality but less accurate than LSTM.  

**Final Recommendation:** Use **LSTM** for stock forecasting.  

## Usage
Clone this repo and run the notebook to reproduce results:

```bash
git clone https://huggingface.co/amlucky/DataSynthis_ML_JobTask

##  License
MIT License