Summary

The article discusses the application of the K-means clustering algorithm to identify support and resistance levels in stock price analysis, specifically for Apple's stock (AAPL) from 2013 to 2020.

Abstract

The article published on Towards Data Science explores the use of K-means clustering to determine support and resistance channels in technical analysis of stock prices. It begins with an explanation of support and resistance as price barriers where the price tends to "bounce" within a channel. The author then describes the K-means algorithm's ability to segment time series data into groups, with the highest and lowest values of each group serving as potential support and resistance levels. The Python code provided accesses Apple stock data, optimizes the K-value for clustering, and visualizes the resulting support and resistance lines. The article emphasizes the importance of finding the optimal K-value, which is done by identifying the "elbow point" in the sum of squared distances plot. The final implementation plots the support and resistance lines alongside the stock prices, with each cluster's data points color-coded accordingly. The author acknowledges limitations, such as the subjectivity of support and resistance lines and the potential inaccuracy of the final resistance line.

Opinions

The author believes that support and resistance levels are crucial for understanding entry and exit points in stock trading.
It is suggested that the K-means clustering method can provide a more objective approach to identifying support and resistance levels compared to traditional, subjective methods.
The article posits that the elbow method is an effective way to determine the optimal number of clusters (K-value) for the K-means algorithm.
The author indicates that the final resistance line may be less accurate due to its reliance on the last value in the dataset.
There is an acknowledgment that support and resistance strategies have downsides, including their limited period of effectiveness and susceptibility to human error.
The author provides a personal link for readers to access more of their content, implying a level of expertise and ongoing interest in the subject matter.

Using K-means Clustering to Create Support and Resistance:

Note from Towards Data Science’s editors: While we allow independent authors to publish articles in accordance with our rules and guidelines, we do not endorse each author’s contribution. You should not rely on an author’s works without seeking professional advice. See our Reader Terms for details.

Support and resistance are some of the most talked-about concepts when it comes to technical analysis. Support and resistance are used as price barriers, in which the price “bounces” off of. In this article, I will use the K-means clustering algorithm to find these different support and resistance channels, and trade with these insights.

Support and Resistance:

To understand how best to implement something, we should first understand the thing that we want to implement.

Support and Resistance, are two lines that are drawn on a graph, to form a channel, in which the price exists within.

Support and resistance are resultant of a security not being able to decrease or increase anymore, due to pressure from sellers or buyers. A good rule of thumb is that the more times a price is deflected against a support or resistance line, the less likely it will work again.

Support and resistance give good insight into entry points and selling points, as the support and resistance lines are theoretically the lowest and highest points for that limited time period.

Downsides of the support and resistance strategy is that it works for an unknown period of time, and the lines are subjective and are therefore subject to human error.

Program Concept:

The K-means clustering algorithm, finds different sections of the time series data, and groups them into a defined number of groups. This number (K) can be optimized. The highest and lowest value of each group is then defined as the support and resistance values for the cluster.

Now that we know how the program is intended, let’s try to recreate it in Python!

The Code:

import yfinance
df = yfinance.download('AAPL','2013-1-1','2020-1-1')
X = np.array(df['Close'])

This script is to access data for the Apple stock price. For this example, we are implementing the support and resistance only on the closing price.

from sklearn.cluster import KMeans
import numpy as np
from kneed import DataGenerator, KneeLocator
    
sum_of_squared_distances = []
K = range(1,15)
for k in K:
    km = KMeans(n_clusters=k)
    km = km.fit(X.reshape(-1,1))
    sum_of_squared_distances.append(km.inertia_)
kn = KneeLocator(K, sum_of_squared_distances,S=1.0, curve="convex", direction="decreasing")
kn.plot_knee()
# plt.plot(sum_of_squared_distances)

This script is to test the different values of K to find the best value:

The K-value of 2 creates support and resistance lines that will never be reached for a long time.

A K-value of 9 creates support and resistance that are far too common and make it difficult to make predictions.

Therefore, we have to find the best value of K, calculated by the elbow point when comparing variance between K values. The elbow point is the biggest improvement, given a certain movement.

Based on the kneed library, the elbow point is at 4. This means that the optimum K value is 4.

kmeans = KMeans(n_clusters= kn.knee).fit(X.reshape(-1,1))
c = kmeans.predict(X.reshape(-1,1))
minmax = []
for i in range(kn.knee):
    minmax.append([-np.inf,np.inf])
for i in range(len(X)):
    cluster = c[i]
    if X[i] > minmax[cluster][0]:
        minmax[cluster][0] = X[i]
    if X[i] < minmax[cluster][1]:
        minmax[cluster][1] = X[i]

This script finds the minimum and maximum value for the points that reside in each cluster. These, when plotted, become the support and resistance lines.

from matplotlib import pyplot as plt
for i in range(len(X)):
    colors = ['b','g','r','c','m','y','k','w']
    c = kmeans.predict(X[i].reshape(-1,1))[0]
    color = colors[c]
    plt.scatter(i,X[i],c = color,s = 1)

for i in range(len(minmax)):
    plt.hlines(minmax[i][0],xmin = 0,xmax = len(X),colors = 'g')
    plt.hlines(minmax[i][1],xmin = 0,xmax = len(X),colors = 'r')

This script plots the support and resistance, along with the actual graph of the prices, which are color coded based on the cluster. Unfortunately, I think that the colors are limited, meaning that there is a limited K value in which the data can be color coded.

This is the result of the program, a set of support and resistance lines. Keep in mind that the lines are most accurate, when the values fall back into the channel. Additionally, the final resistance line would be the least accurate ,as it takes the last value into account, without considering any other values.

My links:

If you want to see more of my content, click this link.