Analysis of Singapore’s MRT/LRT passenger flows

This article uses network science to explore passenger flow patterns in Singapore’s MRT/LRT system and evaluates which stations, and segments matter most.

1. Introduction

With the completion of the Circle Line (CCL) loop on 12 July 2026, Singapore’s MRT/LRT network currently spans just under 280km and will be progressively expanded to about 360km by the early 2030s. When completed, Singapore’s total rail length will exceed that of major cities such as Tokyo or Hong Kong today and will be comparable to that of London and New York¹.

On an average weekday, this network successfully carries more than 3.6 million passenger journeys². Demand is largely bimodal with the morning peak from 7am to 9am, and a broader evening peak between 5pm to 8pm.

Network science provides us with a useful set of tools to study this complex system and understand the patterns behind these flows. The graph model was built earlier conceptualising stations as nodes, tracks as edges, weighted by distance, journey time, and observed passenger flows from LTA’s tap-in/tap-out records in Aug 2026.

This article seeks to address three main questions: (a) where do passengers come from and where are they going, (b) which parts of the network carry this movement, and (c) which segments are most important in the network?

2. Where do passengers come from, and where are they going?

2.1 The daily rhythm

Daily passenger journeys have bimodal peaks. The morning peak (7-9 AM) is more compressed, likely a consequence of fixed school and office start times. The evening peak (5-8 PM) is heavier, and lasts longer over three hours instead of two. Between them sits a midday trough at roughly a third of peak volume. This pattern is broadly consistent across all MRT lines.

Figure 1: Average weekday journeys by hour of day. Use the dropdown to toggle journeys for each line.

2.2 Where do journeys begin and end?

Parsing the origin-destination data from LTA allows us to count the total entries and exits at each station during the AM and PM peak windows.

AM Peak (7–9am)

Stations with the most tap-ins and tap-outs, average weekday AM peak (07:00–08:59)

	Busiest origins			Busiest destinations
	Station	Lines	Tap-ins	Station	Lines	Tap-outs
1	Yishun	NS	15,226	Raffles Place	EW NS	29,811
2	Admiralty	NS	14,156	Tanjong Pagar	EW	17,109
3	Woodlands	NS TE	14,152	Jurong East	EW NS	16,794
4	Bukit Panjang	BP DT	14,031	Novena	NS	13,573
5	Sengkang	NE SK	13,982	Outram Park	EW NE TE	12,620
6	Tampines	DT EW	13,963	Downtown	DT	12,435
7	Serangoon	CC NE	13,893	Kent Ridge	CC	11,446
8	Sembawang	NS	13,642	Orchard	NS TE	11,175
9	Lakeside	EW	12,499	Newton	DT NS	10,820
10	Khatib	NS	12,255	Bugis	DT EW	10,591

PM Peak (5–8pm)

Stations with the most tap-ins and tap-outs, average weekday PM peak (17:00–19:59)

	Busiest origins			Busiest destinations
	Station	Lines	Tap-ins	Station	Lines	Tap-outs
1	Raffles Place	EW NS	36,838	Woodlands	NS TE	24,102
2	Jurong East	EW NS	28,141	Yishun	NS	22,201
3	Tanjong Pagar	EW	21,311	Serangoon	CC NE	21,964
4	Orchard	NS TE	19,071	Tampines	DT EW	21,702
5	Bugis	DT EW	18,379	Boon Lay	EW	19,413
6	HarbourFront	CC NE	17,557	Sengkang	NE SK	18,834
7	Novena	NS	16,849	Jurong East	EW NS	18,595
8	Newton	DT NS	16,124	Newton	DT NS	16,100
9	Outram Park	EW NE TE	15,991	Sembawang	NS	16,078
10	City Hall	EW NS	15,620	Bukit Panjang	BP DT	15,853

Unsurprisingly, the busiest AM Peak journeys begin from residential towns such as Yishun, Admiralty, Woodlands and Sembawang along the North-South Line, with Sengkang, Tampines, Bukit Panjang and Lakeside following closely. Most AM Peak journeys end in the central business district and surrounding areas.

Passengers originate relatively evenly from residential areas, but flow is concentrated toward a handful of city centre destinations. Yishun, the top origin station, sends 15.2k passengers, only 24% more than tenth-placed Khatib (12.3k). In contrast, Raffles Place receives 29.8k passengers, 1.7 times the next highest destination station (Tanjong Pagar, 17.1k). Two destinations (Jurong East and Kent Ridge) are outside the centre but draw heavy AM Peak arrivals as a second commercial hub and a university-and-science-park cluster, consistent with decentralisation counter-flows.

By evening the pattern reverses. Eight of the ten busiest AM Peak destinations become the busiest PM Peak origins, with residential towns the top destinations. Jurong East and Newton are among the top 10 PM Peak origins and destinations, presenting themselves as two-way hubs rather than pure workplaces or residential areas.

2.3 How might we characterise stations?

Plotting tap-ins / tap-outs illustrates (a) how busy a station is (total flows) and (b) whether the station sends or receives more people during AM and PM peak hours.

The spatial distribution of origin (sending) and destination (receiving) stations is visible in Figure 2. During AM Peak, the direction of travel is mainly toward the central areas, with some stations like Newton and Clementi having large but balanced bidirectional flows. During PM Peak, the pattern flips, as evidenced by the color swaps with passengers mainly moving out of the central areas toward residential towns. Newton, Paya Lebar and Clementi have large but balanced bidirectional flows.

AM peak (7–9am)

Figure 2: MRT stations by total flows (bubble area: tap-ins plus tap-outs) and direction (colour), average weekday AM peak. Net origins send at least 30% more passengers than they receive; net destinations receive at least 30% more than they send; balanced stations fall in between. The three busiest stations of each category are labelled.

PM peak (5–8pm)

Figure 3: The same stations in the evening peak, on the same scale and with the same thresholds.

3. Which stations and segments are the busiest?

3.1 Assigning flows and calculating throughput

Origin–destination records tell us where each journey starts and ends, but not the route taken in between. To fill this gap, we assume every passenger takes the fastest way — working it out using Dijkstra’s algorithm. Each journey is credited to every segment along its route, and to the station where it changes lines. Add these journeys up gives us the number of people along each segment.

To assess the ‘busyness’ of a station, we measure overall throughput — tap-ins, tap-outs and transfers, which reflects the number of people passing through. This has implications on platform crowding, safety and station design.

3.2 Busiest stations during AM / PM Peak

Looking at the data, interchange stations are the busiest, with transfers accounting for a large share of total passenger flows. The busiest interchanges have intersecting major lines (i.e. North-South (NS) and East-West (EW) lines).

Jurong East stands out with the largest total station throughput, reflecting its critical role in distributing passengers in Western Singapore.

Circle line (CCL) interchange stations also features prominently, reflecting the line’s orbital role in connecting passengers between lines.

• AM Peak: Bishan, Serangoon, Paya Lebar and Buona Vista
• PM Peak: Serangoon, Dhoby Ghaut and Paya Lebar

Thomson-East Coast (TEL) line interchange stations (e.g. Orchard, Woodlands) see relatively less transfers, but could grow in importance with emerging residential areas or as passengers become more familiar with transferring via these stations.

AM Peak (7–9am)

Busiest MRT stations by throughput, average weekday AM peak (07:00–08:59). Tap-ins and tap-outs come from the fare records; transfers are inferred from fastest-path routing and are an upper bound.

	Station	Lines	Tap-ins	Tap-outs	Transfers	Total	% Trf
1	Jurong East	EW NS	6,263	16,794	52,603	75,660	70%
2	Outram Park	EW NE TE	3,035	12,620	50,209	65,865	76%
3	Raffles Place	EW NS	775	29,811	17,847	48,432	37%
4	Bishan	CC NS	10,538	3,483	30,198	44,219	68%
5	Serangoon	CC NE	13,893	4,902	25,273	44,068	57%
6	Paya Lebar	CC EW	5,046	8,898	20,994	34,937	60%
7	Buona Vista	CC EW	2,833	10,128	20,796	33,757	62%
8	Bukit Panjang	BP DT	14,031	5,658	12,903	32,593	40%
9	Dhoby Ghaut	CC NE NS	1,940	5,408	24,252	31,600	77%
10	Woodlands	NS TE	14,152	7,506	9,564	31,222	31%

PM Peak (5–8pm)

The same measure in the evening peak (17:00–19:59).

	Station	Lines	Tap-ins	Tap-outs	Transfers	Total	% Trf
1	Jurong East	EW NS	28,141	18,595	68,645	115,381	59%
2	Outram Park	EW NE TE	15,991	4,713	76,001	96,705	79%
3	City Hall	EW NS	15,620	10,516	46,213	72,349	64%
4	Serangoon	CC NE	11,284	21,964	32,324	65,572	49%
5	Dhoby Ghaut	CC NE NS	10,833	7,777	41,764	60,373	69%
6	Paya Lebar	CC EW	15,502	13,507	31,312	60,322	52%
7	Orchard	NS TE	19,071	13,211	26,640	58,921	45%
8	Raffles Place	EW NS	36,838	5,669	9,500	52,007	18%
9	Bishan	CC NS	7,229	7,562	34,387	49,178	70%
10	Woodlands	NS TE	11,565	24,102	12,815	48,482	26%

3.3 Busiest line segments

Segments along the North-East line consistently feature the most passengers.

During AM Peak, flows between Hougang to Kovan and Kovan to Serangoon are the highest before some passengers exit at Serangoon to switch to other lines. Nonetheless, overall load remains heavy through toward the city (i.e Woodleigh –> Potong Pasir –> Boon Keng –> Farrer Park –> Little India) before dissipating. More than 60,000 passengers travel through the busiest segment (Kovan to Serangoon), roughly equivalent to the capacity of 440 double decker buses!

The picture is consistent during PM peak, with heavy passenger flows away from the city on the North-East line. The Buona Vista–Dover–Clementi stretch also carries heavy evening flows westward, swelling at Buona Vista as passengers transfer from the Circle Line.

AM Peak (7–9am)

Figure 4: Track segments by peak-direction load, average weekday AM peak. Line width shows busier direction’s volume.

Busiest track segments by peak-direction load, average weekday AM peak (07:00–08:59). Direction shown is the busier of the two; peak share is that direction’s load as a fraction of the two-way total.

	From	To	Line	Peak direction	Reverse	Peak share
1	Kovan	Serangoon	NE	60,141	8,007	88%
2	Hougang	Kovan	NE	56,837	7,121	89%
3	Farrer Park	Little India	NE	56,159	7,359	88%
4	Boon Keng	Farrer Park	NE	55,144	6,848	89%
5	Woodleigh	Potong Pasir	NE	54,987	7,163	88%
6	Potong Pasir	Boon Keng	NE	54,809	7,005	89%
7	Serangoon	Woodleigh	NE	54,439	7,386	88%
8	Clementi	Dover	EW	53,265	17,860	75%
9	Ang Mo Kio	Bishan	NS	53,180	10,658	83%
10	Little India	Dhoby Ghaut	NE	52,880	7,187	88%

PM Peak (5–8pm)

Figure 5: The same measure in the evening peak (17:00–19:59), on the same scale.

The same measure in the evening peak (17:00–19:59).

	From	To	Line	Peak direction	Reverse	Peak share
1	Little India	Farrer Park	NE	67,925	18,099	79%
2	Serangoon	Kovan	NE	67,100	17,355	79%
3	Dhoby Ghaut	Little India	NE	66,691	18,764	78%
4	Boon Keng	Potong Pasir	NE	64,474	16,591	80%
5	Potong Pasir	Woodleigh	NE	64,247	16,757	79%
6	Farrer Park	Boon Keng	NE	64,041	17,452	79%
7	Dover	Clementi	EW	63,624	32,243	66%
8	Buona Vista	Dover	EW	63,223	33,067	66%
9	Woodleigh	Serangoon	NE	63,042	17,130	79%
10	Kovan	Hougang	NE	62,328	15,866	80%

4. Which parts of the network are most important?

While Sections 2 and 3 show which stations and segments have the highest volume, passenger traffic alone does not define system criticality. A low-volume station may serve as a single point of failure between two major hubs, while a high-traffic station might have close substitutes. Network science provides several measures to evaluate topological importance and identify components that maintain system connectivity.

4.1 An overview of centrality measures

Betweenness centrality counts how often a station lies on the path between two others. A station with high betweenness is a broker: passengers pass through it on their way somewhere else, and if it closed, their journeys would have to be rerouted. Interchanges tend to score highly, but so do stations at structural pinch points, where a single stretch of track is the only connection between two parts of the network.

Closeness centrality measures how near a station is to everywhere else, on average. A station with high closeness is well placed — journeys from it reach the rest of the network quickly.

Both measures are built on shortest paths, so the results depend on how “shortest” is defined and which journeys are counted. We compute each three ways:

1. Topological structure. Every stretch of track is equal, and the measures reflect the shape of the network.
2. Weighted by travel time. The shortest path is the quickest one. Transfer penalties at interchanges are taken into account.
3. Weighted by passenger demand. For betweenness, each pair of stations is counted by the number of journeys actually made between them, so a station scores highly when it lies on routes many people travel. Closeness has no meaningful demand-weighted form and hence we do not report the computations.

4.2 What the rankings show

Topological network. On the unweighted network, eight of the ten stations with the highest betweenness centrality are Circle Line interchanges (Buona Vista, Botanic Gardens, Bishan, Serangoon, MacPherson, Caldecott, Paya Lebar and Marina Bay). The Circle Line is orbital, crossing every radial line. In a network where every stop counts equally it offers the shortest way between any two lines without passing through the city centre. Outram Park tops the list as the meeting point of three lines on the edge of the central business district. Closeness tells a different story where the best-placed stations (Little India, Newton, Dhoby Ghaut, Rochor and Bugis) sit around the geographic centre of the network, a similar number of stops from everywhere.

Travel time weighted. Once paths are measured in minutes, stations near the city centre rise in ranking. Several Circle Line interchanges fall away from the top of the betweenness ranking. Serangoon and Bishan fall in rank, as routes that save a stop but cost a transfer lose their advantage. Chinatown climbs thirteen places. The closeness ranking shifts more significantly, the geographic centre gives way to the southern CBD, and three Thomson–East Coast Line stations with no interchange (Havelock, Maxwell and Great World) climb twenty or more places. On the newest line, running directly through the heart of the city, a station can reach much of the network quickly without changing trains. Dhoby Ghaut, despite serving three lines, slips five places. An interchange gives options, but come with a cost in transfer time. As such, a station near several interchanges, on a direct line, can outperform one that is itself an interchange.

Passenger demand weighted. Counting the journeys people actually make, Outram Park remains first, and Serangoon, Bishan and Buona Vista stay near the top. However, the rest of the list now features stations along corridors identified in Section 3. Dhoby Ghaut rises eleven places, City Hall twelve, and two ordinary stops with no interchange at all, Dover and Clementi, climb ten places each. They have no structural reason to rank highly, but sit on the single East-West Line corridor carrying the west’s AM Peak commute into the city, so a large share of all journeys pass through them. Closeness is not computed for this layer.

Topological

Top MRT stations by betweenness and closeness on the unweighted network, where every stretch of track counts as one step.

	Betweenness			Closeness
	Station	Lines		Station	Lines
1	Outram Park	EW NE TE		Little India	DT NE
2	Buona Vista	CC EW		Newton	DT NS
3	Botanic Gardens	CC DT		Dhoby Ghaut	CC NE NS
4	Bishan	CC NS		Rochor	DT
5	Serangoon	CC NE		Bugis	DT EW
6	MacPherson	CC DT		Stevens	DT TE
7	Caldecott	CC TE		City Hall	EW NS
8	Paya Lebar	CC EW		Botanic Gardens	CC DT
9	Marina Bay	CC NS TE		Clarke Quay	NE
10	Little India	DT NE		Caldecott	CC TE

Travel time

The same measures with paths weighted by journey time, including interchange walking time and wait penalty. Change is the movement in rank from the structural layer.

	Betweenness			Closeness
	Station	Lines	Change	Station	Lines	Change
1	Outram Park	EW NE TE	–	Outram Park	EW NE TE	▲ 10
2	Buona Vista	CC EW	–	Havelock	TE	▲ 20
3	MacPherson	CC DT	▲ 3	Chinatown	DT NE	▲ 9
4	Chinatown	DT NE	▲ 13	Maxwell	TE	▲ 28
5	Paya Lebar	CC EW	▲ 3	Great World	TE	▲ 25
6	Bishan	CC NS	▼ 2	City Hall	EW NS	▲ 1
7	Marina Bay	CC NS TE	▲ 2	Tanjong Pagar	EW	▲ 12
8	Jurong East	EW NS	▲ 6	Dhoby Ghaut	CC NE NS	▼ 5
9	Serangoon	CC NE	▼ 4	Orchard	NS TE	▲ 5
10	Stevens	DT TE	▲ 2	Raffles Place	EW NS	▲ 3

Passenger demand

Betweenness with each station pair weighted by its journeys, average weekday AM peak. Closeness is not shown as it has no meaningful demand-weighted form. Change is the movement in rank from the travel-time layer.

	Betweenness
	Station	Lines	Change
1	Outram Park	EW NE TE	–
2	Serangoon	CC NE	▲ 7
3	Buona Vista	CC EW	▼ 1
4	Dhoby Ghaut	CC NE NS	▲ 11
5	Bishan	CC NS	▲ 1
6	Jurong East	EW NS	▲ 2
7	Little India	DT NE	▲ 7
8	Dover	EW	▲ 10
9	City Hall	EW NS	▲ 12
10	Clementi	EW	▲ 10

4.3 Three kinds of importance

The three layers broadly separate stations into three groups.

• Important on every measure. Outram Park, Buona Vista, Bishan and Serangoon rank highly whether the network is judged by its shape, its speed or its usage. These are the stations where structural position and actual demand coincide, and where any disruption would be felt most widely.
• Important by position, less intensely used. Botanic Gardens, Caldecott, MacPherson, Paya Lebar and Marina Bay, which are all Circle Line interchanges, matter in the structure of the network but drop out once demand is counted. They represent orbital connections the network has but does not yet fully rely on during the AM peak, routes that could absorb passengers diverted from a disrupted radial line.
• Important by use, not position. Dover and Clementi rank lower on structure but high on demand. Their importance comes from the volume funnelled through them rather than from any unique place in the network. A corridor that carries heavy demand with few nearby alternatives is where a disruption would cause the most difficulty.

The contrast between the first and last groups is notable. While topology alone places emphasis on interchanges, actual demand shows that some of the stations the network relies on most are ordinary stops on busy single-line corridors.

MRT stations ranking in the top 10 on at least 3 of the five centrality measures. Btw = betweenness, Cls = closeness. Closeness has no demand-weighted form, so the maximum possible total is five.

		Structure		Travel time		Demand
Station	Lines	Btw	Cls	Btw	Cls	Btw	Total
Outram Park	EW NE TE	✓		✓	✓	✓	4
Serangoon	CC NE	✓		✓		✓	3
Buona Vista	CC EW	✓		✓		✓	3
Dhoby Ghaut	CC NE NS		✓		✓	✓	3
Bishan	CC NS	✓		✓		✓	3
Little India	DT NE	✓	✓			✓	3
City Hall	EW NS		✓		✓	✓	3

5. Conclusion

My overall take is that the numbers mostly confirm what anyone who rides the MRT already feels. Each morning, passengers pour out of residential towns and converge on a handful of stations in the city centre, and each evening the whole pattern runs in reverse. The North East Line carries the heaviest of that tide, with trains far fuller in one direction than the other.

What surprised me most was how differently each station looks depending on how you measure its network importance. Measured by topology alone, the Circle Line interchanges stand out as the stations that knit everything together. Measured by how people actually travel, the spotlight moves to the busy radial corridors, including stations like Dover and Clementi that matter simply because so many journeys pass through them. Outram Park ranks near the top on almost every measure, with Buona Vista, Bishan and Serangoon close behind as the network’s most consistent brokers. These stations merit close attention from a resilience and reliability perspective.

There are limits to all this. The model assumes everyone takes the fastest route and is based on monthly aggregated data. But it does raise a question worth exploring next - what happens to these journeys when one of these key stations is disrupted? With the Circle Line loop now complete, it would also be interesting to see whether those orbital connections start carrying more of the load in the months ahead.

This article was written in tandem with coursework for the CASA0002 Urban Simulation module.

References

Barabási, A.-L. (2025) Network Science by Albert-László Barabási. Available at: http://networksciencebook.com/ (Accessed: September 10, 2026).

CNA (2026) “Blending form and function: A first look at 3 new MRT stations that will close the Circle Line loop - CNA,” Channel News Asia. Available at: https://www.channelnewsasia.com/singapore/circle-line-cantonment-price-edward-road-keppel-mrt-stations-first-look-6172836 (Accessed: September 10, 2026).

GeekforGeeks (2026) “Find Shortest Paths from Source to all Vertices using Dijkstra’s Algorithm,” GeeksforGeeks. Available at: https://www.geeksforgeeks.org/dsa/dijkstras-shortest-path-algorithm-greedy-algo-7/ (Accessed: September 10, 2026).

LTA (2015) “Train Station Chinese Names LTA data.gov.sg.” Available at: https://data.gov.sg/datasets/d_d312a5b127e1ae74299b8ae664cedd4e/view (Accessed: September 10, 2026).

LTA (2026) “LTA Dynamic Datasets.” Available at: https://datamall.lta.gov.sg/content/datamall/en/dynamic-data.html#Public%20Transport (Accessed: September 10, 2026).

NetworkX (2025) “NetworkX — NetworkX documentation.” Available at: https://networkx.org/en/ (Accessed: September 10, 2026).

OSM (2026) “Singapore/Rail Transport - OpenStreetMap Wiki.” Available at: https://wiki.openstreetmap.org/wiki/Singapore/Rail_Transport (Accessed: September 10, 2026).

URA (2026) “Master Plan 2025 Planning Area Boundary (No Sea) URA.” Available at: https://data.gov.sg/datasets/d_2cc750190544007400b2cfd5d7f53209/view (Accessed: September 10, 2026).

Footnotes

1. Growing Singapore’s land transport network (LTA, 2026)

2. More train rides taken in first half-year, but overall public transport use stays below 2019 levels, (ST, 2025)